技术领域technical field
本公开涉及图像处理技术领域，尤其涉及一种图像处理方法及装置。The present disclosure relates to the technical field of image processing, and in particular, to an image processing method and device.
背景技术Background technique
随着通信技术的发展，手机的使用越来越广泛，用户可以使用手机拨打语音电话，通过手机安装的即时通信软件发送文字或图片，还可以使用手机上网浏览图片并进行保存。为了保证图片发送速率以及减小图片存储时占用的存储空间，手机通常会对图片进行压缩，常用的图片压缩格式包括JPEG(Joint Photographic Experts Group，联合图像专家小组)，BMP(Bitmap)，PNG(Portable Network Graphic Format，图像文件存储格式)以及GIF(Graphics Interchange Format，图像互换格式)等。With the development of communication technology, the use of mobile phones is becoming more and more extensive. Users can use mobile phones to make voice calls, send text or pictures through instant messaging software installed on mobile phones, and also use mobile phones to browse pictures online and save them. In order to ensure the picture transmission rate and reduce the storage space occupied by the picture storage, the mobile phone usually compresses the picture. Commonly used picture compression formats include JPEG (Joint Photographic Experts Group, Joint Photographic Experts Group), BMP (Bitmap), PNG ( Portable Network Graphic Format, image file storage format) and GIF (Graphics Interchange Format, image interchange format), etc.
发明内容Contents of the invention
为克服相关技术中存在的问题，本公开实施例提供一种图像处理方法及装置。所述技术方案如下：In order to overcome the problems existing in related technologies, embodiments of the present disclosure provide an image processing method and device. Described technical scheme is as follows:
根据本公开实施例的第一方面，提供一种图像处理方法，包括：According to a first aspect of an embodiment of the present disclosure, an image processing method is provided, including:
获取待显示图像的灰度图和色度图；Obtain the grayscale and chromaticity images of the image to be displayed;
获取所述灰度图的骨架图和纹理图；Obtain the skeleton image and texture image of the grayscale image;
分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图；Repairing the skeleton map and the texture map respectively, and obtaining the repaired skeleton map and the repaired texture map;
根据所述修复骨架图和修复纹理图，获取修复灰度图；Obtain a repaired grayscale image according to the repaired skeleton map and the repaired texture map;
根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。Displaying the restored image of the image to be displayed according to the repaired grayscale image and the chromaticity image.
本公开的实施例提供的技术方案可以包括以下有益效果：通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。The technical solutions provided by the embodiments of the present disclosure may include the following beneficial effects: by respectively repairing the skeleton map and the texture map of the grayscale image to be displayed, the repaired skeleton map and the repaired texture map are obtained, and then the repaired skeleton map and the repaired texture map are obtained. The image and the chromaticity image of the image to be displayed are synthesized into a restored image and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space. Thereby improving the user experience.
在一个实施例中，所述获取所述灰度图的骨架图和纹理图包括：In one embodiment, said obtaining the skeleton image and the texture image of the grayscale image comprises:
通过全变分算法将所述灰度图分解为所述骨架图和所述纹理图。The grayscale image is decomposed into the skeleton image and the texture image through a total variation algorithm.
本公开的实施例提供的技术方案可以包括以下有益效果：通过全变分算法将灰度图分解为骨架图和纹理图，提高了灰度图分解的效率，同时提高了骨架图和纹理图的准确率。The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: the grayscale image is decomposed into a skeleton image and a texture image through a full variational algorithm, which improves the efficiency of grayscale image decomposition and improves the efficiency of the skeleton image and texture image at the same time. Accuracy.
在一个实施例中，所述分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图包括：In one embodiment, the said skeleton map and texture map are respectively repaired, and obtaining the repaired skeleton map and repaired texture map includes:
通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图；Repairing the skeleton diagram through the first full convolutional network to obtain the repaired skeleton diagram;
通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图。The texture map is repaired through the second full convolutional network to obtain the repaired texture map.
本公开的实施例提供的技术方案可以包括以下有益效果：分别通过第一全卷积网络和第二全卷积网络对骨架图和纹理图进行修复，提高了修复骨架图和纹理图的效率，同时提高了修复骨架图和纹理图的精确度。The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: the skeleton map and the texture map are respectively repaired through the first full convolution network and the second full convolution network, which improves the efficiency of repairing the skeleton map and the texture map, At the same time, the accuracy of repairing skeleton maps and texture maps is improved.
在一个实施例中，所述通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图包括：In one embodiment, the repairing the skeleton diagram through the first full convolutional network, and obtaining the repairing skeleton diagram includes:
通过所述第一全卷积网络的第一卷积层对所述骨架图进行特征提取，获取特征骨架图，所述第一卷积层的卷积核为3*3，步长为1；Carry out feature extraction to described skeleton figure by the first convolutional layer of described first full convolutional network, obtain characteristic skeleton figure, the convolution kernel of described first convolutional layer is 3*3, step size is 1;
通过所述第一全卷积网络的第二卷积层对所述特征骨架图进行降维，获取降维骨架图，所述第二卷积层的卷积核为3*3，步长为2；Through the second convolutional layer of the first full convolutional network, the feature skeleton diagram is reduced in dimension, and the dimensionality reduction skeleton diagram is obtained. The convolution kernel of the second convolutional layer is 3*3, and the step size is 2;
通过所述第一全卷积网络的第三卷积层对所述降维骨架图进行特征映射，获取映射骨架图，所述第三卷积层的卷积核为1*1，步长为2；Through the third convolutional layer of the first full convolutional network, the dimensionality reduction skeleton diagram is subjected to feature mapping to obtain a mapped skeleton diagram, the convolution kernel of the third convolutional layer is 1*1, and the step size is 2;
通过所述第一全卷积网络的第四卷积层对所述映射骨架图进行升维采样，获取所述修复骨架图，所述第四卷积层为反卷积层，卷积核为3*3，步长为2。Through the fourth convolutional layer of the first full convolutional network, the mapping skeleton is sampled to obtain the repaired skeleton, the fourth convolutional layer is a deconvolution layer, and the convolution kernel is 3*3, with a step size of 2.
本公开的实施例提供的技术方案可以包括以下有益效果：在通过第一全卷积网络对骨架图进行修复时，通过不同卷积层的计算，在获取修复骨架图的同时，保证了骨架图和修复骨架图之间的尺寸一致，提高了骨架图修复的精确度。The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: when the skeleton diagram is repaired through the first full convolutional network, the calculation of different convolution layers can obtain the repaired skeleton diagram while ensuring the skeleton diagram The size is consistent with the repaired skeleton diagram, which improves the accuracy of skeleton diagram repair.
在一个实施例中，所述通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图包括：In one embodiment, the repairing the texture map through the second full convolutional network, and obtaining the repaired texture map includes:
通过所述第二全卷积网络的第五卷积层对所述纹理图进行特征提取，获取特征纹理图，所述第五卷积层的卷积核为3*3，步长为1；Feature extraction is performed on the texture map through the fifth convolutional layer of the second full convolutional network to obtain a feature texture map, the convolution kernel of the fifth convolutional layer is 3*3, and the step size is 1;
通过所述第二全卷积网络的第六卷积层对所述特征纹理图进行降维，获取降维纹理图，所述第六卷积层的卷积核为3*3，步长为2；Through the sixth convolutional layer of the second full convolutional network, the feature texture map is reduced in dimension, and the dimensionality reduction texture map is obtained. The convolution kernel of the sixth convolutional layer is 3*3, and the step size is: 2;
通过所述第二全卷积网络的第七卷积层对所述降维纹理图进行特征映射，获取映射纹理图，所述第七卷积层的卷积核为1*1，步长为2；Perform feature mapping on the dimensionality reduction texture map through the seventh convolutional layer of the second full convolutional network to obtain a mapped texture map, the convolution kernel of the seventh convolutional layer is 1*1, and the step size is 2;
通过所述第二全卷积网络的第八卷积层对所述映射纹理图进行升维采样，获取所述修复纹理图，所述第八卷积层为反卷积层，卷积核为3*3，步长为2。The eighth convolutional layer of the second full convolutional network is used to upscale the mapped texture map to obtain the repaired texture map, the eighth convolutional layer is a deconvolution layer, and the convolution kernel is 3*3, with a step size of 2.
本公开的实施例提供的技术方案可以包括以下有益效果：在通过第二全卷积网络对纹理图进行修复时，通过不同卷积层的计算，在获取修复纹理图的同时，保证了纹理图和修复纹理图之间的尺寸一致，提高了纹理图修复的精确度。The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: when repairing the texture map through the second full convolutional network, through the calculation of different convolution layers, while obtaining the repaired texture map, the texture map is guaranteed. The size is consistent with the repaired texture map, which improves the accuracy of texture map repair.
根据本公开实施例的第二方面，提供一种图像处理装置，包括：According to a second aspect of an embodiment of the present disclosure, an image processing device is provided, including:
第一获取模块，用于获取待显示图像的灰度图和色度图；The first obtaining module is used to obtain the grayscale image and the chromaticity image of the image to be displayed;
第二获取模块，用于获取所述灰度图的骨架图和纹理图；The second acquisition module is used to acquire the skeleton image and the texture image of the grayscale image;
修复模块，用于分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图；A repair module, configured to repair the skeleton map and the texture map respectively, and obtain the repaired skeleton map and the repaired texture map;
第三获取模块，用于根据所述修复骨架图和修复纹理图，获取修复灰度图；The third obtaining module is used to obtain the repaired grayscale image according to the repaired skeleton map and the repaired texture map;
显示模块，用于根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。A display module, configured to display the restored image of the image to be displayed according to the repaired grayscale image and the chromaticity image.
在一个实施例中，所述第二获取模块包括：In one embodiment, the second acquisition module includes:
第一获取子模块，用于通过全变分算法将所述灰度图分解为所述骨架图和所述纹理图。The first acquisition sub-module is configured to decompose the grayscale image into the skeleton image and the texture image through a full variational algorithm.
在一个实施例中，所述修复模块包括：In one embodiment, the repair module includes:
第一修复子模块，用于通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图；The first repair submodule is used to repair the skeleton diagram through the first full convolutional network, and obtain the repaired skeleton diagram;
第二修复子模块，用于通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图。The second repairing sub-module is used to repair the texture map through the second full convolutional network, and obtain the repaired texture map.
在一个实施例中，所述第一修复子模块包括：In one embodiment, the first repair submodule includes:
第一修复单元，用于通过所述第一全卷积网络的第一卷积层对所述骨架图进行特征提取，获取特征骨架图，所述第一卷积层的卷积核为3*3，步长为1；The first repair unit is used to perform feature extraction on the skeleton map through the first convolutional layer of the first full convolutional network to obtain a feature skeleton map, and the convolution kernel of the first convolutional layer is 3* 3, the step size is 1;
第二修复单元，用于通过所述第一全卷积网络的第二卷积层对所述特征骨架图进行降维，获取降维骨架图，所述第二卷积层的卷积核为3*3，步长为2；The second repair unit is used to reduce the dimensionality of the feature skeleton diagram through the second convolutional layer of the first full convolutional network to obtain a dimensionality-reduced skeleton diagram, and the convolution kernel of the second convolutional layer is: 3*3, the step size is 2;
第三修复单元，用于通过所述第一全卷积网络的第三卷积层对所述降维骨架图进行特征映射，获取映射骨架图，所述第三卷积层的卷积核为1*1，步长为2；The third restoration unit is used to perform feature mapping on the dimensionality reduction skeleton diagram through the third convolutional layer of the first full convolutional network to obtain a mapped skeleton diagram, and the convolution kernel of the third convolutional layer is: 1*1, the step size is 2;
第四修复单元，用于通过所述第一全卷积网络的第四卷积层对所述映射骨架图进行升维采样，获取所述修复骨架图，所述第四卷积层为反卷积层，卷积核为3*3，步长为2。The fourth repairing unit is used to perform up-dimensional sampling on the mapped skeleton diagram through the fourth convolutional layer of the first full convolutional network to obtain the repaired skeleton diagram, and the fourth convolutional layer is deconvolution Multilayer, the convolution kernel is 3*3, and the step size is 2.
在一个实施例中，所述第二修复子模块包括：In one embodiment, the second repair submodule includes:
第五修复单元，用于通过所述第二全卷积网络的第五卷积层对所述纹理图进行特征提取，获取特征纹理图，所述第五卷积层的卷积核为3*3，步长为1；The fifth restoration unit is used to perform feature extraction on the texture map through the fifth convolutional layer of the second full convolutional network to obtain a characteristic texture map, and the convolution kernel of the fifth convolutional layer is 3* 3, the step size is 1;
第六修复单元，用于通过所述第二全卷积网络的第六卷积层对所述特征纹理图进行降维，获取降维纹理图，所述第六卷积层的卷积核为3*3，步长为2；The sixth repair unit is used to reduce the dimensionality of the feature texture map through the sixth convolutional layer of the second full convolutional network to obtain the dimensionality reduction texture map, and the convolution kernel of the sixth convolutional layer is 3*3, the step size is 2;
第七修复单元，用于通过所述第二全卷积网络的第七卷积层对所述降维纹理图进行特征映射，获取映射纹理图，所述第七卷积层的卷积核为1*1，步长为2；The seventh repair unit is configured to perform feature mapping on the dimensionality reduction texture map through the seventh convolutional layer of the second full convolutional network to obtain a mapped texture map, and the convolution kernel of the seventh convolutional layer is 1*1, the step size is 2;
第八修复单元，用于通过所述第二全卷积网络的第八卷积层对所述映射纹理图进行升维采样，获取所述修复纹理图，所述第八卷积层为反卷积层，卷积核为3*3，步长为2。The eighth repair unit is configured to perform up-dimensional sampling on the mapped texture map through the eighth convolutional layer of the second full convolutional network to obtain the repaired texture map, and the eighth convolutional layer is deconvolution Multilayer, the convolution kernel is 3*3, and the step size is 2.
根据本公开实施例的第三方面，提供一种图像处理装置，包括：According to a third aspect of an embodiment of the present disclosure, an image processing device is provided, including:
处理器；processor;
用于存储处理器可执行指令的存储器；memory for storing processor-executable instructions;
其中，所述处理器被配置为：Wherein, the processor is configured as:
获取待显示图像的灰度图和色度图；Obtain the grayscale and chromaticity images of the image to be displayed;
获取所述灰度图的骨架图和纹理图；Obtain the skeleton image and texture image of the grayscale image;
分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图；Repairing the skeleton map and the texture map respectively, and obtaining the repaired skeleton map and the repaired texture map;
根据所述修复骨架图和修复纹理图，获取修复灰度图；Obtain a repaired grayscale image according to the repaired skeleton map and the repaired texture map;
根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。Displaying the restored image of the image to be displayed according to the repaired grayscale image and the chromaticity image.
根据本公开实施例的第四方面，提供一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现第一方面任一实施例所述方法的步骤。According to a fourth aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in any embodiment of the first aspect are implemented.
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本公开。It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.
附图说明Description of drawings
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本公开的实施例，并与说明书一起用于解释本公开的原理。The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description serve to explain the principles of the disclosure.
图1是根据一示例性实施例示出的有损压缩后的图像的示意图。Fig. 1 is a schematic diagram of a lossy compressed image according to an exemplary embodiment.
图2a是根据一示例性实施例示出的图像处理方法的流程图。Fig. 2a is a flowchart of an image processing method according to an exemplary embodiment.
图2b是根据一示例性实施例示出的修复后的图像的示意图。Fig. 2b is a schematic diagram of a repaired image according to an exemplary embodiment.
图2c是根据一示例性实施例示出的原始图像的示意图。Fig. 2c is a schematic diagram of an original image according to an exemplary embodiment.
图2d是根据一示例性实施例示出的图像处理方法的流程图。Fig. 2d is a flowchart of an image processing method according to an exemplary embodiment.
图3是根据一示例性实施例示出的图像处理方法的流程图。Fig. 3 is a flowchart of an image processing method according to an exemplary embodiment.
图4a是根据一示例性实施例示出的图像处理装置的结构示意图。Fig. 4a is a schematic structural diagram of an image processing device according to an exemplary embodiment.
图4b是根据一示例性实施例示出的图像处理装置的结构示意图。Fig. 4b is a schematic structural diagram of an image processing device according to an exemplary embodiment.
图4c是根据一示例性实施例示出的图像处理装置的结构示意图。Fig. 4c is a schematic structural diagram of an image processing device according to an exemplary embodiment.
图4d是根据一示例性实施例示出的图像处理装置的结构示意图。Fig. 4d is a schematic structural diagram of an image processing device according to an exemplary embodiment.
图4e是根据一示例性实施例示出的图像处理装置的结构示意图。Fig. 4e is a schematic structural diagram of an image processing device according to an exemplary embodiment.
图5是根据一示例性实施例示出的图像处理装置的结构框图。Fig. 5 is a structural block diagram of an image processing device according to an exemplary embodiment.
具体实施方式detailed description
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本公开的一些方面相一致的装置和方法的例子。Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present disclosure as recited in the appended claims.
本公开实施例提供的技术方案涉及终端，该终端可以为手机，平板电脑，个人电脑以及其他能够显示图像的设备，本公开实施例对此不作限定。相关技术中，为了保证图片发送速率以及减小图片存储时占用的存储空间，终端通常会对图片进行压缩，常用的图片压缩格式包括JPEG，BMP，PNG以及GIF等。但是这些压缩格式均为有损压缩，如果压缩率过大，会损失图像中的有用信息，对图像的清晰度造成影响，例如JPEG图片中的块效应，振铃效应和模糊效应等。如图1所示，图1为经过JPEG格式压缩过的图片，从图1中可以明显看出图片中的块效应，导致图像的清晰度较差，进而导致用户查看图片时的体验较差。本公开的实施例提供的技术方案中，通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为待显示图像的复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。The technical solution provided by the embodiments of the present disclosure relates to a terminal, and the terminal may be a mobile phone, a tablet computer, a personal computer or other devices capable of displaying images, which is not limited in the embodiments of the present disclosure. In related technologies, in order to ensure the picture transmission rate and reduce the storage space occupied by the pictures, the terminal usually compresses the pictures, and commonly used picture compression formats include JPEG, BMP, PNG, and GIF. However, these compression formats are all lossy compression. If the compression rate is too high, the useful information in the image will be lost, which will affect the clarity of the image, such as block effects, ringing effects and blur effects in JPEG pictures. As shown in Figure 1, Figure 1 is a picture compressed in JPEG format. From Figure 1, it can be clearly seen that the block effect in the picture leads to poor image clarity, which in turn leads to poor user experience when viewing pictures. In the technical solution provided by the embodiments of the present disclosure, by respectively repairing the skeleton map and the texture map of the grayscale image to be displayed, the repaired skeleton map and the repaired texture map are obtained, and then the repaired skeleton map, the repaired texture map and the image to be displayed The chromaticity diagram of the image is synthesized into a restored image of the image to be displayed and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space. Thereby improving the user experience.
图2a是根据一示例性实施例示出的一种图像处理方法的流程图，该方法用于终端，该终端可以为手机，平板电脑，个人电脑以及其他能够显示图像的设备。如图2a所示，该图像处理方法包括以下步骤201至步骤205：Fig. 2a is a flow chart of an image processing method according to an exemplary embodiment, the method is used in a terminal, and the terminal may be a mobile phone, a tablet computer, a personal computer and other devices capable of displaying images. As shown in Figure 2a, the image processing method includes the following steps 201 to 205:
在步骤201中，获取待显示图像的灰度图和色度图。In step 201, a grayscale image and a chromaticity image of an image to be displayed are obtained.
示例的，用户在浏览网页时，可以根据需要选择图像进行查看，假设该网页上设置有多个图像，用户可以点击第一图像，即此时用户需要查看第一图像，终端确认接收到用户的该次点击之后，可以获取网页上的第一图像的下载地址，然后根据该下载地址下载第一图像。此时下载得到的第一图像为有损压缩后的图像，如果直接显示可能会出现清晰度较差的情况，因此终端可以首先对该第一图像进行修复。For example, when a user browses a web page, he can select an image to view as needed. Assuming that there are multiple images set on the web page, the user can click on the first image, that is, the user needs to view the first image at this time, and the terminal confirms that it has received the user's After the click, the download address of the first image on the webpage can be obtained, and then the first image can be downloaded according to the download address. At this time, the downloaded first image is a lossy compressed image, and if it is directly displayed, the definition may be poor, so the terminal may repair the first image first.
具体的，第一图像即为待显示图像，终端可以首先将待显示图像从RGB(Red-Green-Blue，红绿蓝)色彩空间转到YUV(Luminance-Chrominance，灰度色度)色彩空间，然后获取转换后Y通道的图像作为待显示图像的灰度图，获取转换后UV通道的图像作为待显示图像的色度图。Specifically, the first image is the image to be displayed, and the terminal may first convert the image to be displayed from the RGB (Red-Green-Blue, red-green-blue) color space to the YUV (Luminance-Chrominance, grayscale chroma) color space, Then the image of the converted Y channel is obtained as the grayscale image of the image to be displayed, and the image of the converted UV channel is obtained as the chromaticity image of the image to be displayed.
在步骤202中，获取灰度图的骨架图和纹理图。In step 202, a skeleton image and a texture image of the grayscale image are acquired.
示例的，终端可以采用全变分算法将灰度图分解为骨架图和纹理图。例如，终端可以搭建全变分算法模型，然后将灰度图中每个像素的灰度输入至全变分算法模型，即输入全变分算法模型的为体现灰度图中每个像素灰度的分段平滑函数，经过全变分算法模型的运算，即可输出该灰度图的骨架图和纹理图。通过全变分算法将灰度图分解为骨架图和纹理图，提高了灰度图分解的效率，同时提高了骨架图和纹理图的准确率。For example, the terminal may use a full variation algorithm to decompose the grayscale image into a skeleton image and a texture image. For example, the terminal can build a full variational algorithm model, and then input the grayscale of each pixel in the grayscale image to the full variational algorithm model, that is, the input of the full variational algorithm model is to reflect the grayscale of each pixel in the grayscale image. The piecewise smoothing function of , after the operation of the full variational algorithm model, the skeleton map and texture map of the grayscale image can be output. The grayscale image is decomposed into a skeleton image and a texture image through a total variation algorithm, which improves the efficiency of the gray image decomposition and improves the accuracy of the skeleton image and the texture image.
在步骤203中，分别对骨架图和纹理图进行修复，获取修复骨架图和修复纹理图。In step 203, the skeleton map and the texture map are repaired respectively, and the repaired skeleton map and the repaired texture map are obtained.
示例的，终端可以通过全卷积网络分别对骨架图和纹理图进行修复，获取修复骨架图和修复纹理图。For example, the terminal may respectively repair the skeleton map and the texture map through the fully convolutional network, and obtain the repaired skeleton map and the repaired texture map.
通常的全卷积网络由多个卷积层和全连接层组成，由于全连接层要求固定的输入大小，为了使得该全卷积网络能够修复不同尺寸的图像，本公开实施例所述的全卷积网络可以去掉全连接层，便于对任意尺寸的图像进行处理，且经过设计可以保证全卷积网络的输入的图像和输出的图像具有相同的尺寸。The usual fully convolutional network consists of multiple convolutional layers and fully connected layers. Since the fully connected layer requires a fixed input size, in order to enable the fully convolutional network to repair images of different sizes, the fully convolutional network described in the embodiments of the present disclosure The convolutional network can remove the fully connected layer, which is convenient for processing images of any size, and is designed to ensure that the input image and the output image of the full convolutional network have the same size.
在步骤204中，根据修复骨架图和修复纹理图，获取修复灰度图。In step 204, according to the repaired skeleton map and the repaired texture map, the repaired grayscale image is obtained.
示例的，终端在获取到修复骨架图和修复纹理图之后，可以根据全变分的逆算法将修复骨架图和修复纹理图合成为修复灰度图。For example, after acquiring the repaired skeleton map and the repaired texture map, the terminal may synthesize the repaired skeleton map and the repaired texture map into a repaired grayscale image according to a total variation inverse algorithm.
在步骤205中，根据修复灰度图和色度图显示待显示图像的复原图像。In step 205, the restored image of the image to be displayed is displayed according to the restored grayscale image and the chromaticity image.
示例的，终端可以将待显示图像的修复灰度图和待显示图像的原始色度图合成为待显示图像的复原图像，并将该复原图像从YUV色彩空间转换至RGB色彩空间，然后在显示屏上显示该转换后的复原图像，终端显示的复原图像的清晰度远远大于待显示图像。例如，参考图1所示，有损压缩后的待显示图像的块效应明显，而经过修复后终端显示的复原图像如图2b所示，由图2b可知复原图像中的块效应明显减小，即清晰度得到了显著提高。同时，与图2c所示的原始图像相比，可以明显看出图2b所示的复原图像与图2c所示的原始图像的清晰度更接近，即复原图像更精确的反应了原始图像的信息。For example, the terminal can synthesize the restored grayscale image of the image to be displayed and the original chromaticity image of the image to be displayed into a restored image of the image to be displayed, and convert the restored image from the YUV color space to the RGB color space, and then display The converted restored image is displayed on the screen, and the definition of the restored image displayed on the terminal is far greater than that of the image to be displayed. For example, referring to FIG. 1, the image to be displayed after lossy compression has obvious blockiness, and the restored image displayed by the terminal after restoration is shown in FIG. 2b. It can be seen from FIG. 2b that the blockiness in the restored image is significantly reduced. That is, sharpness is significantly improved. At the same time, compared with the original image shown in Figure 2c, it can be clearly seen that the restored image shown in Figure 2b is closer to the definition of the original image shown in Figure 2c, that is, the restored image more accurately reflects the information of the original image .
本公开的实施例提供的技术方案中，通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。In the technical solution provided by the embodiments of the present disclosure, by respectively repairing the skeleton map and the texture map of the grayscale image to be displayed, the repaired skeleton map and the repaired texture map are obtained, and then the repaired skeleton map, the repaired texture map and the image to be displayed The chromaticity map of the image is synthesized into a restored image and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space, thereby improving user experience. experience.
在一个实施例中，如图2d所示，在步骤203中，分别对骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，可以通过步骤2031和步骤2032实现：In one embodiment, as shown in FIG. 2d, in step 203, the skeleton map and the texture map are respectively repaired, and the repair skeleton map and the repair texture map are obtained, which can be realized through steps 2031 and 2032:
在步骤2031中，通过第一全卷积网络对骨架图进行修复，获取修复骨架图。In step 2031, the skeleton graph is repaired through the first full convolutional network to obtain the repaired skeleton graph.
示例的，初始化时，终端可以搭建图像修复模型，该图像修复模型中包括用于修复骨架图的第一全卷积网络和用于修复纹理图的第二全卷积网络，然后将多个原始图像已知的训练图像输入图像修复模型，获取通过图像修复模型修复得到的多个训练图像的复原图像，通过不断调整第一全卷积网络和第二全卷积网络的参数，使得经过图像修复模型修复得到的训练图像的复原图像与训练图像的原始图像之间的差异小于或等于预设阈值，此时说明图像修复模型训练完成。实际应用中，也可以采用同一个全卷积网络同时对骨架图和纹理图进行修复，即终端搭建的图像修复模型中仅包括用于同时修复骨架图和纹理图的一个全卷积网络。For example, during initialization, the terminal can build an image inpainting model, the image inpainting model includes a first fully convolutional network for inpainting a skeleton map and a second full convolutional network for inpainting a texture map, and then multiple original The known training image of the image is input into the image inpainting model, and the restored images of multiple training images repaired by the image inpainting model are obtained, and the parameters of the first full convolutional network and the second full convolutional network are continuously adjusted, so that after image inpainting If the difference between the restored image of the training image obtained by model inpainting and the original image of the training image is less than or equal to a preset threshold, it indicates that the training of the image inpainting model is completed. In practical applications, the same full convolutional network can also be used to repair the skeleton map and the texture map at the same time, that is, the image repair model built by the terminal only includes a full convolutional network for simultaneously repairing the skeleton map and the texture map.
在训练时，可以将无压缩的原始图像进行不同程度的压缩，得到不同程度的压缩后的图像，然后将压缩后的图像和其对应的原始图像，作为训练样本。During training, the uncompressed original image can be compressed to different degrees to obtain compressed images of different degrees, and then the compressed image and its corresponding original image can be used as training samples.
具体的，终端在通过第一全卷积网络对骨架图进行修复，获取修复骨架图时，可以首先通过第一全卷积网络的卷积核为3*3，步长为1的第一卷积层对骨架图进行特征提取，获取特征骨架图，然后通过第一全卷积网络的卷积核为3*3，步长为2的第二卷积层对特征骨架图进行降维，同时进行特征增强，获取到降维骨架图，接着通过第一全卷积网络的卷积核为1*1，步长为2的第三卷积层对降维骨架图进行特征映射，获取映射骨架图，最后通过第一全卷积网络的卷积核为3*3，步长为2的反卷积层对映射骨架图进行升维采样，获取修复骨架图，该反卷积层即为第一全卷积网络的第四卷积层。通过不同卷积层的计算，在获取修复骨架图的同时，保证了骨架图和修复骨架图之间的尺寸一致，提高了骨架图修复的精确度。Specifically, when the terminal repairs the skeleton diagram through the first full convolutional network and obtains the repaired skeleton diagram, it can first pass through the first volume of the first full convolutional network with a convolution kernel of 3*3 and a step size of 1. The multilayer performs feature extraction on the skeleton map to obtain the feature skeleton map, and then reduces the dimensionality of the feature skeleton map through the second convolution layer of the first full convolutional network with a convolution kernel of 3*3 and a step size of 2. Perform feature enhancement to obtain a dimensionality reduction skeleton diagram, and then perform feature mapping on the dimensionality reduction skeleton diagram through the third convolutional layer with a convolution kernel of 1*1 and a step size of 2 in the first full convolutional network to obtain a mapping skeleton Finally, through the deconvolution layer of the first full convolutional network with a convolution kernel of 3*3 and a step size of 2, the mapped skeleton image is upscaled and sampled to obtain a repaired skeleton image. This deconvolution layer is the first The fourth convolutional layer of a fully convolutional network. Through the calculation of different convolutional layers, while obtaining the repaired skeleton map, the size consistency between the skeleton map and the repaired skeleton map is ensured, and the accuracy of the skeleton map repair is improved.
在步骤2032中，通过第二全卷积网络对纹理图进行修复，获取修复纹理图。In step 2032, the texture map is repaired through the second full convolutional network to obtain the repaired texture map.
示例的，终端在通过第二全卷积网络对纹理图进行修复，获取修复纹理图时，可以首先通过第二全卷积网络的卷积核为3*3，步长为1的第五卷积层对纹理图进行特征提取，获取特征纹理图，然后通过第二全卷积网络的卷积核为3*3，步长为2的第六卷积层对特征纹理图进行降维，同时进行特征增强，获取降维纹理图，接着通过第二全卷积网络的卷积核为1*1，步长为2第七卷积层对降维纹理图进行特征映射，获取映射纹理图，最后通过第二全卷积网络的卷积核为3*3，步长为2的反卷积层对映射纹理图进行升维采样，获取修复纹理图，该反卷积层即为第二全卷积网络的第八卷积层。通过不同卷积层的计算，在获取修复纹理图的同时，保证了纹理图和修复纹理图之间的尺寸一致，提高了纹理图修复的精确度。For example, when the terminal repairs the texture map through the second full convolutional network and obtains the repaired texture map, it can first pass through the fifth volume of the second full convolutional network with a convolution kernel of 3*3 and a step size of 1. The accumulation layer performs feature extraction on the texture map, obtains the feature texture map, and then reduces the dimensionality of the feature texture map through the sixth convolution layer with a convolution kernel of 3*3 and a step size of 2 in the second full convolutional network, and at the same time Perform feature enhancement to obtain a dimensionality reduction texture map, and then pass through the second full convolution network with a convolution kernel of 1*1 and a step size of 2. The seventh convolutional layer performs feature mapping on the dimensionality reduction texture map to obtain a mapped texture map, Finally, through the deconvolution layer of the second full convolutional network with a convolution kernel of 3*3 and a step size of 2, the mapped texture map is upscaled and sampled to obtain the repaired texture map. This deconvolution layer is the second full convolutional layer. The eighth convolutional layer of the convolutional network. Through the calculation of different convolutional layers, while obtaining the repaired texture map, the size consistency between the texture map and the repaired texture map is ensured, and the accuracy of the texture map repair is improved.
本公开的实施例提供的技术方案中，分别通过第一全卷积网络和第二全卷积网络对骨架图和纹理图进行修复，提高了修复骨架图和纹理图的效率，同时提高了修复骨架图和纹理图的精确度。In the technical solution provided by the embodiments of the present disclosure, the skeleton map and the texture map are respectively repaired through the first full convolution network and the second full convolution network, which improves the efficiency of repairing the skeleton map and the texture map, and improves the restoration efficiency. Accuracy of skeleton and texture maps.
下面通过几个实施例详细介绍实现过程。The implementation process will be described in detail below through several embodiments.
图3是根据一示例性实施例示出的一种图像处理方法的流程图，执行主体为终端，如图3所示，包括以下步骤301至步骤314：Fig. 3 is a flow chart of an image processing method according to an exemplary embodiment, where the execution subject is a terminal, as shown in Fig. 3 , including the following steps 301 to 314:
在步骤301中，将待显示图像从RGB色彩空间转换至YUV色彩空间，获取待显示YUV图像。In step 301, the image to be displayed is converted from the RGB color space to the YUV color space, and the YUV image to be displayed is obtained.
在步骤302中，获取该待显示YUV图像的Y通道的灰度图和UV通道的色度图。In step 302, the grayscale image of the Y channel and the chromaticity image of the UV channel of the YUV image to be displayed are acquired.
在步骤303中，通过全变分算法将该灰度图分解为骨架图和纹理图。In step 303, the grayscale image is decomposed into a skeleton image and a texture image through a total variation algorithm.
在步骤304中，通过第一全卷积网络的卷积核为3*3，步长为1的第一卷积层对该骨架图进行特征提取，获取特征骨架图。In step 304, feature extraction is performed on the skeleton map through the first full convolutional network with a convolution kernel of 3*3 and a step size of 1 to obtain a feature skeleton map.
在步骤305中，通过第一全卷积网络的卷积核为3*3，步长为2的第二卷积层对特征骨架图进行降维，获取降维骨架图。In step 305, the dimensionality of the feature skeleton is reduced by the second convolution layer of the first full convolutional network with a convolution kernel of 3*3 and a step size of 2 to obtain a dimensionality-reduced skeleton.
在步骤306中，通过第一全卷积网络的卷积核为1*1，步长为2的第三卷积层对降维骨架图进行特征映射，获取映射骨架图；In step 306, the dimensionality reduction skeleton diagram is subjected to feature mapping through the third convolution layer whose convolution kernel is 1*1 and the step size is 2 to obtain the mapped skeleton diagram;
在步骤307中，通过第一全卷积网络的卷积核为3*3，步长为2的反卷积层对映射骨架图进行升维采样，获取修复骨架图。In step 307, the deconvolution layer with a convolution kernel of 3*3 and a step size of 2 in the first full convolutional network performs upsampling of the mapped skeleton image to obtain the repaired skeleton image.
在步骤308中，通过第二全卷积网络的卷积核为3*3，步长为1的第五卷积层对纹理图进行特征提取，获取特征纹理图。In step 308, feature extraction is performed on the texture map through the fifth convolution layer of the second full convolutional network with a convolution kernel of 3*3 and a step size of 1 to obtain a feature texture map.
在步骤309中，通过第二全卷积网络的卷积核为3*3，步长为2的第六卷积层对特征纹理图进行降维，获取降维纹理图。In step 309, the dimensionality reduction of the feature texture map is performed through the sixth convolution layer of the second full convolutional network with a convolution kernel of 3*3 and a step size of 2, to obtain a dimensionality-reduced texture map.
在步骤310中，通过第二全卷积网络的卷积核为1*1，步长为2的第七卷积层对降维纹理图进行特征映射，获取映射纹理图。In step 310, feature mapping is performed on the dimensionality-reduced texture map through the seventh convolutional layer with a convolution kernel of 1*1 and a step size of 2 in the second full convolutional network to obtain a mapped texture map.
在步骤311中，通过第二全卷积网络的卷积核为3*3，步长为2的反卷积层对映射纹理图进行升维采样，获取修复纹理图。In step 311, the deconvolution layer of the second full convolutional network with a convolution kernel of 3*3 and a step size of 2 performs upsampling of the mapped texture map to obtain the repaired texture map.
在步骤312中，将修复骨架图和修复纹理图合成为修复灰度图。In step 312, the repaired skeleton image and the repaired texture image are synthesized into a repaired grayscale image.
在步骤313中，将修复灰度图和色度图合成为待显示图像的复原图像。In step 313, the repaired grayscale image and the chromaticity image are synthesized into a restored image of the image to be displayed.
在步骤314中，将该复原图像转换至RGB色彩空间，并进行显示。In step 314, the restored image is converted into RGB color space and displayed.
需要说明的时，本公开实施例所述的步骤301至步骤314的顺序在实际应用中可以根据具体情况进行调整，本公开实施例对此不作限定。同时，第一全卷积网络和第二全卷积网络可以为相同的卷积网络也可以为不同的卷积网络，本公开实施例对此不作限定。When it needs to be explained, the sequence of step 301 to step 314 described in the embodiment of the present disclosure may be adjusted according to specific situations in actual application, which is not limited in the embodiment of the present disclosure. Meanwhile, the first full convolutional network and the second full convolutional network may be the same convolutional network or different convolutional networks, which is not limited in this embodiment of the present disclosure.
本公开的实施例提供一种图像处理方法，通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。Embodiments of the present disclosure provide an image processing method. By respectively repairing the skeleton map and the texture map of the grayscale image to be displayed, the repaired skeleton map and the repaired texture map are obtained, and then the repaired skeleton map, the repaired texture map and the texture map to be displayed are obtained. The chromaticity diagram of the displayed image is synthesized into a restored image and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space, thereby improving user experience.
下述为本公开装置实施例，可以用于执行本公开方法实施例。The following are device embodiments of the present disclosure, which can be used to implement the method embodiments of the present disclosure.
图4a是根据一示例性实施例示出的一种图像处理装置40的结构示意图，该装置40可以通过软件、硬件或者两者的结合实现成为电子设备的部分或者全部。如图4a所示，该图像处理装置40包括第一获取模块401，第二获取模块402，修复模块403，第三获取模块404和显示模块405。Fig. 4a is a schematic structural diagram of an image processing apparatus 40 according to an exemplary embodiment. The apparatus 40 may be implemented as part or all of an electronic device through software, hardware or a combination of the two. As shown in FIG. 4 a , the image processing device 40 includes a first acquisition module 401 , a second acquisition module 402 , a restoration module 403 , a third acquisition module 404 and a display module 405 .
其中，第一获取模块401，用于获取待显示图像的灰度图和色度图。Wherein, the first acquiring module 401 is configured to acquire a grayscale image and a chromaticity image of an image to be displayed.
第二获取模块402，用于获取所述灰度图的骨架图和纹理图。The second acquiring module 402 is configured to acquire the skeleton image and the texture image of the grayscale image.
修复模块403，用于分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图。The restoration module 403 is configured to respectively repair the skeleton map and the texture map, and obtain the repaired skeleton map and the repaired texture map.
第三获取模块404，用于根据所述修复骨架图和修复纹理图，获取修复灰度图。The third obtaining module 404 is configured to obtain a repaired grayscale image according to the repaired skeleton map and the repaired texture map.
显示模块405，用于根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。The display module 405 is configured to display the restored image of the image to be displayed according to the restored grayscale image and the chromaticity image.
在一个实施例中，如图4b所示，所述第二获取模块402包括第一获取子模块4021。In one embodiment, as shown in FIG. 4 b , the second acquiring module 402 includes a first acquiring submodule 4021 .
所述第一获取子模块4021，用于通过全变分算法将所述灰度图分解为所述骨架图和所述纹理图。The first acquiring sub-module 4021 is configured to decompose the grayscale image into the skeleton image and the texture image through a total variation algorithm.
在一个实施例中，如图4c所示，所述修复模块403包括第一修复子模块4031和第二修复子模块4032。In one embodiment, as shown in FIG. 4 c , the repair module 403 includes a first repair submodule 4031 and a second repair submodule 4032 .
其中，第一修复子模块4031，用于通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图。Wherein, the first repairing submodule 4031 is configured to repair the skeleton graph through the first full convolutional network, and obtain the repaired skeleton graph.
第二修复子模块4032，用于通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图。The second repairing sub-module 4032 is configured to repair the texture map through the second full convolutional network, and obtain the repaired texture map.
在一个实施例中，如图4d所示，所述第一修复子模块4031包括第一修复单元4031a，第二修复单元4031b，第三修复单元4031c和第四修复单元4031d。In one embodiment, as shown in FIG. 4d, the first repairing submodule 4031 includes a first repairing unit 4031a, a second repairing unit 4031b, a third repairing unit 4031c and a fourth repairing unit 4031d.
其中，第一修复单元4031a，用于通过所述第一全卷积网络的第一卷积层对所述骨架图进行特征提取，获取特征骨架图，所述第一卷积层的卷积核为3*3，步长为1。Wherein, the first restoration unit 4031a is configured to perform feature extraction on the skeleton diagram through the first convolutional layer of the first full convolutional network to obtain a feature skeleton diagram, and the convolution kernel of the first convolutional layer is 3*3, and the step size is 1.
第二修复单元4031b，用于通过所述第一全卷积网络的第二卷积层对所述特征骨架图进行降维，获取降维骨架图，所述第二卷积层的卷积核为3*3，步长为2。The second repair unit 4031b is configured to reduce the dimensionality of the feature skeleton diagram through the second convolutional layer of the first full convolutional network to obtain a dimensionality-reduced skeleton diagram, and the convolution kernel of the second convolutional layer is 3*3, and the step size is 2.
第三修复单元4031c，用于通过所述第一全卷积网络的第三卷积层对所述降维骨架图进行特征映射，获取映射骨架图，所述第三卷积层的卷积核为1*1，步长为2。The third repair unit 4031c is configured to perform feature mapping on the dimensionality reduction skeleton diagram through the third convolutional layer of the first full convolutional network, obtain a mapped skeleton diagram, and the convolution kernel of the third convolutional layer is 1*1, and the step size is 2.
第四修复单元4031d，用于通过所述第一全卷积网络的第四卷积层对所述映射骨架图进行升维采样，获取所述修复骨架图，所述第四卷积层为反卷积层，卷积核为3*3，步长为2。The fourth repairing unit 4031d is configured to perform up-dimensional sampling on the mapped skeleton diagram through the fourth convolutional layer of the first full convolutional network to obtain the repaired skeleton diagram, and the fourth convolutional layer is an inverse Convolution layer, the convolution kernel is 3*3, and the step size is 2.
在一个实施例中，如图4e所示，所述第二修复子模块4032包括第五修复单元4032a，第六修复单元4032b，第七修复单元4032c和第八修复单元4032d。In one embodiment, as shown in FIG. 4e, the second repairing submodule 4032 includes a fifth repairing unit 4032a, a sixth repairing unit 4032b, a seventh repairing unit 4032c and an eighth repairing unit 4032d.
其中，第五修复单元4032a，用于通过所述第二全卷积网络的第五卷积层对所述纹理图进行特征提取，获取特征纹理图，所述第五卷积层的卷积核为3*3，步长为1。Wherein, the fifth repairing unit 4032a is configured to perform feature extraction on the texture map through the fifth convolutional layer of the second full convolutional network to obtain a feature texture map, and the convolution kernel of the fifth convolutional layer is 3*3, and the step size is 1.
第六修复单元4032b，用于通过所述第二全卷积网络的第六卷积层对所述特征纹理图进行降维，获取降维纹理图，所述第六卷积层的卷积核为3*3，步长为2。The sixth repair unit 4032b is configured to reduce the dimensionality of the feature texture map through the sixth convolutional layer of the second full convolutional network, and obtain the reduced dimensionality texture map, and the convolution kernel of the sixth convolutional layer is 3*3, and the step size is 2.
第七修复单元4032c，用于通过所述第二全卷积网络的第七卷积层对所述降维纹理图进行特征映射，获取映射纹理图，所述第七卷积层的卷积核为1*1，步长为2。The seventh repair unit 4032c is configured to perform feature mapping on the dimensionality reduction texture map through the seventh convolutional layer of the second full convolutional network, obtain the mapped texture map, and the convolution kernel of the seventh convolutional layer is 1*1, and the step size is 2.
第八修复单元4032d，用于通过所述第二全卷积网络的第八卷积层对所述映射纹理图进行升维采样，获取所述修复纹理图，所述第八卷积层为反卷积层，卷积核为3*3，步长为2。The eighth repair unit 4032d is configured to perform upsampling on the mapped texture map through the eighth convolutional layer of the second full convolutional network to obtain the repaired texture map, and the eighth convolutional layer is an inverse Convolution layer, the convolution kernel is 3*3, and the step size is 2.
本公开的实施例提供一种图像处理装置，该装置可以通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。Embodiments of the present disclosure provide an image processing device, which can obtain the repaired skeleton map and repaired texture map by respectively repairing the skeleton map and texture map of the grayscale image to be displayed, and then the repaired skeleton map, repaired texture The image and the chromaticity image of the image to be displayed are synthesized into a restored image and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space. Thereby improving the user experience.
本公开实施例提供一种图像处理装置，该图像处理装置包括：An embodiment of the present disclosure provides an image processing device, and the image processing device includes:
处理器；processor;
用于存储处理器可执行指令的存储器；memory for storing processor-executable instructions;
其中，处理器被配置为：where the processor is configured as:
获取待显示图像的灰度图和色度图；Obtain the grayscale and chromaticity images of the image to be displayed;
获取所述灰度图的骨架图和纹理图；Obtain the skeleton image and texture image of the grayscale image;
分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图；Repairing the skeleton map and the texture map respectively, and obtaining the repaired skeleton map and the repaired texture map;
根据所述修复骨架图和修复纹理图，获取修复灰度图；Obtain a repaired grayscale image according to the repaired skeleton map and the repaired texture map;
根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。Displaying the restored image of the image to be displayed according to the repaired grayscale image and the chromaticity image.
在一个实施例中，上述处理器还可被配置为：通过全变分算法将所述灰度图分解为所述骨架图和所述纹理图。In an embodiment, the above-mentioned processor may also be configured to: decompose the grayscale image into the skeleton image and the texture image by using a total variation algorithm.
在一个实施例中，上述处理器还可被配置为：通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图；通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图。In an embodiment, the above-mentioned processor may also be configured to: repair the skeleton map through the first full convolutional network to obtain the repaired skeleton map; repair the texture map through the second full convolutional network Repair, get the repair texture map.
在一个实施例中，上述处理器还可被配置为：通过所述第一全卷积网络的第一卷积层对所述骨架图进行特征提取，获取特征骨架图，所述第一卷积层的卷积核为3*3，步长为1；通过所述第一全卷积网络的第二卷积层对所述特征骨架图进行降维，获取降维骨架图，所述第二卷积层的卷积核为3*3，步长为2；通过所述第一全卷积网络的第三卷积层对所述降维骨架图进行特征映射，获取映射骨架图，所述第三卷积层的卷积核为1*1，步长为2；通过所述第一全卷积网络的第四卷积层对所述映射骨架图进行升维采样，获取所述修复骨架图，所述第四卷积层为反卷积层，卷积核为3*3，步长为2。In one embodiment, the above-mentioned processor can also be configured to: perform feature extraction on the skeleton map through the first convolutional layer of the first full convolutional network to obtain a feature skeleton map, and the first convolutional network The convolution kernel of the layer is 3*3, and the step size is 1; through the second convolutional layer of the first full convolutional network, the dimensionality reduction is performed on the feature skeleton diagram to obtain the dimensionality reduction skeleton diagram, and the second The convolution kernel of the convolutional layer is 3*3, and the step size is 2; the dimensionality reduction skeleton diagram is subjected to feature mapping through the third convolutional layer of the first full convolutional network, and the mapped skeleton diagram is obtained, and the The convolution kernel of the third convolutional layer is 1*1, and the step size is 2; through the fourth convolutional layer of the first full convolutional network, the mapping skeleton is sampled to obtain the repaired skeleton. In the figure, the fourth convolution layer is a deconvolution layer, the convolution kernel is 3*3, and the step size is 2.
在一个实施例中，上述处理器还可被配置为：通过所述第二全卷积网络的第五卷积层对所述纹理图进行特征提取，获取特征纹理图，所述第五卷积层的卷积核为3*3，步长为1；通过所述第二全卷积网络的第六卷积层对所述特征纹理图进行降维，获取降维纹理图，所述第六卷积层的卷积核为3*3，步长为2；通过所述第二全卷积网络的第七卷积层对所述降维纹理图进行特征映射，获取映射纹理图，所述第七卷积层的卷积核为1*1，步长为2；通过所述第二全卷积网络的第八卷积层对所述映射纹理图进行升维采样，获取所述修复纹理图，所述第八卷积层为反卷积层，卷积核为3*3，步长为2。In one embodiment, the above-mentioned processor can also be configured to: perform feature extraction on the texture map through the fifth convolutional layer of the second full convolutional network to obtain a feature texture map, and the fifth convolutional network The convolution kernel of the layer is 3*3, and the step size is 1; through the sixth convolutional layer of the second full convolutional network, the feature texture map is reduced in dimension, and the dimensionality reduction texture map is obtained, and the sixth The convolution kernel of the convolutional layer is 3*3, and the step size is 2; the dimensionality reduction texture map is subjected to feature mapping through the seventh convolutional layer of the second full convolutional network to obtain a mapped texture map, and the The convolution kernel of the seventh convolutional layer is 1*1, and the step size is 2; through the eighth convolutional layer of the second full convolutional network, the mapped texture map is upscaled and sampled to obtain the repaired texture In the figure, the eighth convolutional layer is a deconvolutional layer, the convolution kernel is 3*3, and the step size is 2.
本公开的实施例提供一种图像处理装置，该装置可以通过分别对待显示图像灰度图的骨架图和纹理图进行修复，获取修复骨架图和修复纹理图，进而将该修复骨架图、修复纹理图以及待显示图像的色度图合成为复原图像并进行显示，降低了有损压缩对图像显示的影响，在不增加网络传输带宽和硬盘存储空间的前提下，提高了图像显示的清晰度，进而提高了用户体验。Embodiments of the present disclosure provide an image processing device, which can obtain the repaired skeleton map and repaired texture map by respectively repairing the skeleton map and texture map of the grayscale image to be displayed, and then the repaired skeleton map, repaired texture The image and the chromaticity image of the image to be displayed are synthesized into a restored image and displayed, which reduces the impact of lossy compression on image display, and improves the clarity of image display without increasing network transmission bandwidth and hard disk storage space. Thereby improving the user experience.
关于上述实施例中的装置，其中各个模块执行操作的具体方式已经在有关该方法的实施例中进行了详细描述，此处将不做详细阐述说明。Regarding the apparatus in the foregoing embodiments, the specific manner in which each module executes operations has been described in detail in the embodiments related to the method, and will not be described in detail here.
图5是根据一示例性实施例示出的一种用于图像处理装置50的结构框图，该装置50适用于终端设备。例如，装置50可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等。Fig. 5 is a structural block diagram showing an image processing apparatus 50 according to an exemplary embodiment, and the apparatus 50 is suitable for a terminal device. For example, the device 50 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.
装置50可以包括以下一个或多个组件：处理组件502，存储器504，电源组件506，多媒体组件508，音频组件510，输入/输出(I/O)的接口512，传感器组件514，以及通信组件516。Apparatus 50 may include one or more of the following components: processing component 502, memory 504, power supply component 506, multimedia component 508, audio component 510, input/output (I/O) interface 512, sensor component 514, and communication component 516 .
处理组件502通常控制装置50的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件502可以包括一个或多个处理器520来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件502可以包括一个或多个模块，便于处理组件502和其他组件之间的交互。例如，处理组件502可以包括多媒体模块，以方便多媒体组件508和处理组件502之间的交互。Processing component 502 generally controls the overall operations of device 50, such as those associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 502 may include one or more processors 520 to execute instructions to complete all or part of the steps of the above method. Additionally, processing component 502 may include one or more modules that facilitate interaction between processing component 502 and other components. For example, processing component 502 may include a multimedia module to facilitate interaction between multimedia component 508 and processing component 502 .
存储器504被配置为存储各种类型的数据以支持在装置50的操作。这些数据的示例包括用于在装置50上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器504可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。Memory 504 is configured to store various types of data to support operations at device 50 . Examples of such data include instructions for any application or method operating on device 50, contact data, phonebook data, messages, pictures, videos, and the like. The memory 504 can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
电源组件506为装置50的各种组件提供电力。电源组件506可以包括电源管理系统，一个或多个电源，及其他与为装置50生成、管理和分配电力相关联的组件。The power supply component 506 provides power to the various components of the device 50 . Power components 506 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for device 50 .
多媒体组件508包括在所述装置50和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件508包括一个前置摄像头和/或后置摄像头。当装置50处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。The multimedia component 508 includes a screen that provides an output interface between the device 50 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may not only sense a boundary of a touch or swipe action, but also detect duration and pressure associated with the touch or swipe action. In some embodiments, the multimedia component 508 includes a front camera and/or a rear camera. When the device 50 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capability.
音频组件510被配置为输出和/或输入音频信号。例如，音频组件510包括一个麦克风(MIC)，当装置50处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器504或经由通信组件516发送。在一些实施例中，音频组件510还包括一个扬声器，用于输出音频信号。The audio component 510 is configured to output and/or input audio signals. For example, the audio component 510 includes a microphone (MIC) configured to receive external audio signals when the device 50 is in operation modes, such as call mode, recording mode and voice recognition mode. Received audio signals may be further stored in memory 504 or sent via communication component 516 . In some embodiments, the audio component 510 also includes a speaker for outputting audio signals.
I/O接口512为处理组件502和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。The I/O interface 512 provides an interface between the processing component 502 and a peripheral interface module. The peripheral interface module may be a keyboard, a click wheel, a button, and the like. These buttons may include, but are not limited to: a home button, volume buttons, start button, and lock button.
传感器组件514包括一个或多个传感器，用于为装置50提供各个方面的状态评估。例如，传感器组件514可以检测到装置50的打开/关闭状态，组件的相对定位，例如所述组件为装置50的显示器和小键盘，传感器组件514还可以检测装置50或装置50一个组件的位置改变，用户与装置50接触的存在或不存在，装置50方位或加速/减速和装置50的温度变化。传感器组件514可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件514还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件514还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。Sensor assembly 514 includes one or more sensors for providing various aspects of status assessment for device 50 . For example, the sensor component 514 can detect the open/closed state of the device 50, the relative positioning of components, such as the display and keypad of the device 50, and the sensor component 514 can also detect a change in the position of the device 50 or a component of the device 50 , the presence or absence of user contact with the device 50 , the device 50 orientation or acceleration/deceleration and the temperature change of the device 50 . Sensor assembly 514 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. Sensor assembly 514 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor component 514 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.
通信组件516被配置为便于装置50和其他设备之间有线或无线方式的通信。装置50可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件516经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件516还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。The communication component 516 is configured to facilitate wired or wireless communication between the apparatus 50 and other devices. The device 50 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 516 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 516 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.
在示例性实施例中，装置50可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子组件实现，用于执行上述方法。In an exemplary embodiment, apparatus 50 may be programmed by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation for performing the methods described above.
在示例性实施例中，还提供了一种包括指令的非临时性计算机可读存储介质，例如包括指令的存储器504，上述指令可由装置50的处理器520执行以完成上述方法。例如，所述非临时性计算机可读存储介质可以是ROM、随机存取存储器(RAM)、CD-ROM、磁带、软盘和光数据存储设备等。In an exemplary embodiment, there is also provided a non-transitory computer-readable storage medium including instructions, such as the memory 504 including instructions, which can be executed by the processor 520 of the device 50 to implement the above method. For example, the non-transitory computer readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.
本公开实施例还提供一种非临时性计算机可读存储介质，当所述存储介质中的指令由装置50的处理器执行时，使得装置50能够执行上述图像处理方法，所述方法包括An embodiment of the present disclosure also provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the device 50, the device 50 can execute the above-mentioned image processing method, and the method includes
获取待显示图像的灰度图和色度图；Obtain the grayscale and chromaticity images of the image to be displayed;
获取所述灰度图的骨架图和纹理图；Obtain the skeleton image and texture image of the grayscale image;
分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图；Repairing the skeleton map and the texture map respectively, and obtaining the repaired skeleton map and the repaired texture map;
根据所述修复骨架图和修复纹理图，获取修复灰度图；Obtain a repaired grayscale image according to the repaired skeleton map and the repaired texture map;
根据所述修复灰度图和所述色度图显示所述待显示图像的复原图像。Displaying the restored image of the image to be displayed according to the repaired grayscale image and the chromaticity image.
在一个实施例中，所述获取所述灰度图的骨架图和纹理图包括：通过全变分算法将所述灰度图分解为所述骨架图和所述纹理图。In one embodiment, the acquiring the skeleton image and the texture image of the grayscale image includes: decomposing the grayscale image into the skeleton image and the texture image by using a total variation algorithm.
在一个实施例中，所述分别对所述骨架图和纹理图进行修复，获取修复骨架图和修复纹理图包括：通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图；通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图。In one embodiment, the repairing the skeleton map and the texture map respectively, and obtaining the repaired skeleton map and the repaired texture map include: repairing the skeleton map through a first full convolutional network, and obtaining the repaired skeleton map image; the texture image is repaired by the second full convolutional network to obtain the repaired texture image.
在一个实施例中，所述通过第一全卷积网络对所述骨架图进行修复，获取所述修复骨架图包括：通过所述第一全卷积网络的第一卷积层对所述骨架图进行特征提取，获取特征骨架图，所述第一卷积层的卷积核为3*3，步长为1；通过所述第一全卷积网络的第二卷积层对所述特征骨架图进行降维，获取降维骨架图，所述第二卷积层的卷积核为3*3，步长为2；通过所述第一全卷积网络的第三卷积层对所述降维骨架图进行特征映射，获取映射骨架图，所述第三卷积层的卷积核为1*1，步长为2；通过所述第一全卷积网络的第四卷积层对所述映射骨架图进行升维采样，获取所述修复骨架图，所述第四卷积层为反卷积层，卷积核为3*3，步长为2。In one embodiment, the repairing the skeleton graph through the first full convolutional network, and obtaining the repaired skeleton graph includes: repairing the skeleton graph through the first convolutional layer of the first full convolutional network Feature extraction is performed on the figure to obtain a feature skeleton diagram, the convolution kernel of the first convolutional layer is 3*3, and the step size is 1; the features are processed by the second convolutional layer of the first full convolutional network The skeleton diagram is dimensionally reduced, and the dimensionality reduction skeleton diagram is obtained. The convolution kernel of the second convolutional layer is 3*3, and the step size is 2; Perform feature mapping on the dimensionality reduction skeleton diagram to obtain a mapped skeleton diagram, the convolution kernel of the third convolutional layer is 1*1, and the step size is 2; through the fourth convolutional layer of the first full convolutional network The dimension-up sampling is performed on the mapped skeleton image to obtain the repaired skeleton image, the fourth convolution layer is a deconvolution layer, the convolution kernel is 3*3, and the step size is 2.
在一个实施例中，所述通过第二全卷积网络对所述纹理图进行修复，获取所述修复纹理图包括：通过所述第二全卷积网络的第五卷积层对所述纹理图进行特征提取，获取特征纹理图，所述第五卷积层的卷积核为3*3，步长为1；通过所述第二全卷积网络的第六卷积层对所述特征纹理图进行降维，获取降维纹理图，所述第六卷积层的卷积核为3*3，步长为2；通过所述第二全卷积网络的第七卷积层对所述降维纹理图进行特征映射，获取映射纹理图，所述第七卷积层的卷积核为1*1，步长为2；通过所述第二全卷积网络的第八卷积层对所述映射纹理图进行升维采样，获取所述修复纹理图，所述第八卷积层为反卷积层，卷积核为3*3，步长为2。In one embodiment, the repairing the texture map through the second full convolutional network, and obtaining the repaired texture map includes: repairing the texture through the fifth convolutional layer of the second full convolutional network Feature extraction is performed on the image to obtain a feature texture map, the convolution kernel of the fifth convolutional layer is 3*3, and the step size is 1; the feature is processed by the sixth convolutional layer of the second full convolutional network The texture map is dimensionally reduced to obtain a dimensionally reduced texture map, the convolution kernel of the sixth convolutional layer is 3*3, and the step size is 2; the seventh convolutional layer of the second full convolutional network is used for all Perform feature mapping on the dimensionality reduction texture map to obtain the mapped texture map, the convolution kernel of the seventh convolutional layer is 1*1, and the step size is 2; through the eighth convolutional layer of the second full convolutional network Upsampling the mapped texture image to obtain the repaired texture image, the eighth convolutional layer is a deconvolution layer, the convolution kernel is 3*3, and the step size is 2.
本领域技术人员在考虑说明书及实践这里公开的公开后，将容易想到本公开的其它实施方案。本申请旨在涵盖本公开的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本公开的一般性原理并包括本公开未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本公开的真正范围和精神由下面的权利要求指出。Other embodiments of the present disclosure will be readily apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, and these modifications, uses or adaptations follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not disclosed in the present disclosure . The specification and examples are to be considered exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
应当理解的是，本公开并不局限于上面已经描述并在附图中示出的精确结构，并且可以在不脱离其范围进行各种修改和改变。本公开的范围仅由所附的权利要求来限制。It should be understood that the present disclosure is not limited to the precise constructions which have been described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710399476.0ACN107240076A (en) | 2017-05-31 | 2017-05-31 | Image processing method and device |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710399476.0ACN107240076A (en) | 2017-05-31 | 2017-05-31 | Image processing method and device |
| Publication Number | Publication Date |
|---|---|
| CN107240076Atrue CN107240076A (en) | 2017-10-10 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710399476.0APendingCN107240076A (en) | 2017-05-31 | 2017-05-31 | Image processing method and device |
| Country | Link |
|---|---|
| CN (1) | CN107240076A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109558806A (en)* | 2018-11-07 | 2019-04-02 | 北京科技大学 | The detection method and system of high score Remote Sensing Imagery Change |
| CN109671026A (en)* | 2018-11-28 | 2019-04-23 | 浙江大学 | Gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network |
| CN110738626A (en)* | 2019-10-24 | 2020-01-31 | 广东三维家信息科技有限公司 | Rendering graph optimization method and device and electronic equipment |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105635743A (en)* | 2015-12-30 | 2016-06-01 | 福建师范大学 | Minimum noticeable distortion method and system based on saliency detection and total variation |
| CN106127688A (en)* | 2016-06-30 | 2016-11-16 | 北京大学 | A kind of super-resolution image reconstruction method and system thereof |
| CN106204447A (en)* | 2016-06-30 | 2016-12-07 | 北京大学 | The super resolution ratio reconstruction method with convolutional neural networks is divided based on total variance |
| CN106408519A (en)* | 2015-11-10 | 2017-02-15 | 青岛大学 | Non-local image restoration method based on total variation |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408519A (en)* | 2015-11-10 | 2017-02-15 | 青岛大学 | Non-local image restoration method based on total variation |
| CN105635743A (en)* | 2015-12-30 | 2016-06-01 | 福建师范大学 | Minimum noticeable distortion method and system based on saliency detection and total variation |
| CN106127688A (en)* | 2016-06-30 | 2016-11-16 | 北京大学 | A kind of super-resolution image reconstruction method and system thereof |
| CN106204447A (en)* | 2016-06-30 | 2016-12-07 | 北京大学 | The super resolution ratio reconstruction method with convolutional neural networks is divided based on total variance |
| Title |
|---|
| SUBHAJIT CHAUDHURY 等: "Can fully convolutional networks perform well for general image restoration problems?", 《2017 FIFTEENTH IAPR INTERNATIONAL CONFERENCE ON MACHINE VISION APPLICATIONS (MVA)》* |
| XIAOWEI SHAO 等: "An Image Inpainting Approach Based on the Poisson Equation", 《PROCEEDINGS OF THE SECOND INTERNATIONAL CONFERENCE ON DOCUMENT IMAGE ANALYSIS FOR LIBRARIES (DIAL’06)》* |
| 林云莉 等: "基于图像分解的图像修复技术", 《计算机工程》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109558806A (en)* | 2018-11-07 | 2019-04-02 | 北京科技大学 | The detection method and system of high score Remote Sensing Imagery Change |
| CN109671026A (en)* | 2018-11-28 | 2019-04-23 | 浙江大学 | Gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network |
| CN110738626A (en)* | 2019-10-24 | 2020-01-31 | 广东三维家信息科技有限公司 | Rendering graph optimization method and device and electronic equipment |
| CN110738626B (en)* | 2019-10-24 | 2022-06-28 | 广东三维家信息科技有限公司 | Rendering graph optimization method and device and electronic equipment |
| Publication | Publication Date | Title |
|---|---|---|
| CN107230192B (en) | Image processing method, apparatus, computer-readable storage medium, and mobile terminal | |
| CN106530227B (en) | Image Restoration Method and Device | |
| CN110428378A (en) | Processing method, device and the storage medium of image | |
| CN105225647A (en) | Luminance regulating method and device | |
| CN106778773A (en) | The localization method and device of object in picture | |
| CN110458771A (en) | Image processing method and device, electronic device and storage medium | |
| CN107426088A (en) | Pictorial information processing method and processing device | |
| CN111696039A (en) | Image processing method and device, storage medium and electronic equipment | |
| CN108320263A (en) | A kind of method, device and mobile terminal of image procossing | |
| CN110874809A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN106210446B (en) | Saturation enhancement method and device | |
| CN107240076A (en) | Image processing method and device | |
| CN116017171B (en) | Image processing method, device, electronic equipment, chip and storage medium | |
| CN108198148A (en) | The method and device of image procossing | |
| WO2011003315A1 (en) | Mobile terminal based image processing method and mobile terminal | |
| CN107123078A (en) | The method and device of display image | |
| WO2020078102A1 (en) | Image enhancement method and apparatus, and computer-readable storage medium | |
| CN113393391B (en) | Image enhancement method, image enhancement device, electronic apparatus, and storage medium | |
| US20070104360A1 (en) | System and method for capturing 3D face | |
| CN112508820B (en) | Image processing method and device and electronic equipment | |
| CN112188095B (en) | Photographing method, photographing device and storage medium | |
| CN105976344A (en) | Whiteboard image processing method and whiteboard image processing device | |
| CN113660531A (en) | Video processing method and device, electronic equipment and storage medium | |
| CN107992894A (en) | Image-recognizing method, device and computer-readable recording medium | |
| CN113099038B (en) | Image super-resolution processing method, image super-resolution processing device and storage medium |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20171010 |