CN115512112A - A method of removing background based on color channel operation - Google Patents

Abstract

Translated fromChinese

本发明公开了一种基于颜色通道运算的去背景方法，包括：包括如下步骤：获取待处理图片的样本集；并从中选取一张图片作为原图P；从原图P中分别获取R通道数值r、G通道数值g、B通道数值b；并判断是否存在原图P的掩模图；若存在，获取原图P的掩模图，并据原图P的掩模图，获取原图P的理想掩模图；将原图P的理想掩模图与原图P进行掩模运算，获取原图P的前景；若否则去背景结束。本发明利用彩色原图P中的R通道数值r、G通道数值g、B通道数值b，并且充分利用图片的特征，长分量或相对长分量特征，获取原图P的前景，大大地降低了去背景的计算复杂度，能够有效提升用户体验。

The invention discloses a method for removing the background based on color channel operation, comprising the following steps: obtaining a sample set of pictures to be processed; selecting a picture from it as the original picture P; and obtaining R channel values from the original picture P respectively r, G channel value g, B channel value b; and determine whether there is a mask image of the original image P; if it exists, obtain the mask image of the original image P, and obtain the original image P according to the mask image of the original image P The ideal mask image of the original image P; perform masking operation on the ideal mask image of the original image P and the original image P to obtain the foreground of the original image P; otherwise, go to the background and end. The present invention utilizes the R channel value r, the G channel value g, and the B channel value b in the color original image P, and fully utilizes the characteristics of the picture, long component or relatively long component features, to obtain the prospect of the original image P, greatly reducing the Removing the computational complexity of the background can effectively improve the user experience.

Description

Translated fromChinese

一种基于颜色通道运算的去背景方法A method of removing background based on color channel operation

技术领域technical field

本发明涉及计算机视觉和图像处理领域，尤其涉及一种基于颜色通道运算的去背景方法。The invention relates to the fields of computer vision and image processing, in particular to a method for removing background based on color channel operation.

背景技术Background technique

实时图像识别中去背景方法的计算复杂度直接关系到用户体验。近年来，学术界提出了很多背景去除的方法，有代表性的方法有：门限值法(Threshold Based ImageSegmentation)、边缘分割法(Edge Based Image Segmentation)、模糊数学方法(FuzzyTheory)等。这些方法都是试图定义一种像素分类的方法，达到将前景像素与背景像素分离的目的。抽象点来看，现有的背景去除方案的输入一般是灰度图片，彩色图片也首先会被转换为灰度图片，这个过程丢弃了图片通过色彩来表现的特征，对提高背景去除的正确率和准确率不利；并且现有方法均利用像素之间的灰度值数量关系构造分离方法，例如，计算梯度求得边缘、计算隶属度等，这些计算方法涉及到多轮次的平方与平方根运算，对提高运算速度、提升用户体验不利。另外，现有的方法中没有利用特定应用场景下图片的特点，计算时间长，计算复杂度高。The computational complexity of background removal methods in real-time image recognition is directly related to user experience. In recent years, many background removal methods have been proposed in academia, and the representative methods are: Threshold Based Image Segmentation, Edge Based Image Segmentation, Fuzzy Theory, etc. These methods are trying to define a pixel classification method to achieve the purpose of separating foreground pixels from background pixels. From an abstract point of view, the input of existing background removal schemes is generally a grayscale image, and the color image will first be converted into a grayscale image. This process discards the characteristics of the image through color, which is helpful for improving the accuracy of background removal. And the accuracy rate is unfavorable; and the existing methods all use the gray value quantitative relationship between pixels to construct separation methods, for example, calculate the gradient to obtain the edge, calculate the degree of membership, etc., these calculation methods involve multiple rounds of square and square root operations , which is detrimental to improving computing speed and improving user experience. In addition, the existing methods do not take advantage of the characteristics of pictures in specific application scenarios, and the calculation time is long and the calculation complexity is high.

发明内容Contents of the invention

本发明提供一种基于颜色通道运算的去背景方法，以克服上述技术问题。The present invention provides a background removal method based on color channel operation to overcome the above technical problems.

为了实现上述目的，本发明的技术方案是：In order to achieve the above object, technical scheme of the present invention is:

一种基于颜色通道运算的去背景方法，包括如下步骤：A method for removing background based on color channel operation, comprising the steps of:

S1：获取待处理图片的样本集；并从中选取一张图片作为原图P；S1: Obtain a sample set of pictures to be processed; and select a picture from it as the original picture P;

S2：从所述原图P中分别获取R通道数值r、G通道数值g、B通道数值b；S2: Obtain the value r of the R channel, the value g of the G channel, and the value b of the B channel from the original image P;

S3：根据所述R通道数值r、G通道数值g、B通道数值b，判断是否存在原图P的掩模图；若存在，获取原图P的掩模图，执行S4，否则执行S6；S3: According to the value r of the R channel, the value g of the G channel, and the value b of the B channel, determine whether there is a mask image of the original image P; if it exists, obtain the mask image of the original image P, and execute S4; otherwise, execute S6;

S4：据所述原图P的掩模图，获取原图P的理想掩模图；S4: Obtain an ideal mask image of the original image P according to the mask image of the original image P;

S5：将所述原图P的理想掩模图与所述原图P进行掩模运算，获取原图P的前景；执行S6；S5: Perform a mask operation on the ideal mask image of the original image P and the original image P to obtain the foreground of the original image P; execute S6;

S6：对原图P去背景结束。S6: Remove the background from the original picture P and end.

进一步的，所述S3中，判断是否存在原图P的掩模图方法如下：Further, in said S3, the method of judging whether there is a mask image of the original image P is as follows:

S31：获取所述原图P的长分量图，包括R长分量图、G长分量图和B长分量图；S31: Obtain the long component graph of the original image P, including the R long component graph, the G long component graph, and the B long component graph;

S32：判断所述原图P的长分量图是否为长分量可分图，若原图P的长分量图是长分量可分图，则存在原图P的掩模图，所述长分量可分图即为原图P的掩模图；否则，执行S33；S32: Determine whether the long component graph of the original image P is a long component separable graph, if the long component graph of the original image P is a long component separable graph, then there is a mask image of the original image P, and the long component can be separated The picture is the mask picture of the original picture P; otherwise, execute S33;

S33：获取原图P的相对长分量图；S33: Obtain a relatively long component image of the original image P;

S34：判断所述原图P的相对长分量图是否为相对长分量可分图；若原图P的相对长分量图是相对长分量可分图，则存在原图P的掩模图，所述相对长分量可分图即为原图P的掩模图；若否，则不存在原图P的掩模图。S34: Determine whether the relatively long component graph of the original image P is a relatively long component separable graph; if the relatively long component graph of the original image P is a relatively long component separable graph, then there is a mask graph of the original image P, and the The relatively long component separable graph is the mask graph of the original image P; if not, there is no mask graph of the original image P.

进一步的，所述S31中，获取所述原图P的长分量图如下：Further, in the S31, the long component image of the original image P is obtained as follows:

令原图P＝[C_x,y|C_x,y＝(r,g,b)]，其中，C_x,y是原图P的像素点；则Let the original image P=[C_{x, y} |C_{x, y} = (r, g, b)], where C_{x, y} are the pixels of the original image P; then

式中，DCP_R表示R长分量图；DCP_G表示原图P的G长分量图；DCP_B表示原图P的B长分量图；C_x,y是原图P的像素点；x为像素点横坐标；y均为像素点的纵坐标；r为R通道数值；g为G通道数值；b为B通道数值。In the formula, DCP_R represents the R long component graph; DCP_G represents the G long component graph of the original image P; DCP_B represents the B long component graph of the original image P; C_{x, y} is the pixel point of the original image P; x is the pixel The abscissa of the point; y is the ordinate of the pixel; r is the value of the R channel; g is the value of the G channel; b is the value of the B channel.

进一步的，所述S32中，判断原图P的长分量图是否为长分量可分图的方法如下：Further, in S32, the method for judging whether the long component graph of the original graph P is a long component separable graph is as follows:

若like

为真，则原图P的长分量图为R长分量可分图；If is true, the long component graph of the original graph P is a long component separable graph of R;

若like

为真，则原图P的长分量图为G长分量可分图；is true, the long component graph of the original graph P is a long component separable graph of G;

若like

为真，则原图P的长分量图为B长分量可分图；is true, the long component graph of the original graph P is a separable graph of B long component;

式中，

为存在集合中任意一元素；∧为合取运算符；∈为集合的属于运算符；

为蕴含运算符；O为原图P的前景；B为原图P的前背景。In the formula,

is any element in the existence set; ∧ is the conjunction operator; ∈ is the belonging operator of the set;

is the implication operator; O is the foreground of the original image P; B is the foreground and background of the original image P.

进一步的，所述S33中，获取原图P的相对长分量图如下：Further, in said S33, the relatively long component graph of the original image P is obtained as follows:

所述原图P的相对长分量图包括R对G相对长分量图、R对B相对长分量图、G对R相对长分量图、G对B相对长分量图、B对R相对长分量图、B对G相对长分量图；The relatively long component graphs of the original image P include R to G relatively long component graphs, R to B relatively long component graphs, G to R relatively long component graphs, G to B relatively long component graphs, and B to R relatively long component graphs , B to G relative long component graph;

获取所述G对B相对长分量图如下：Obtain the relative long component graph of G versus B as follows:

获取所述G对R相对长分量图如下：Obtain the relative long component graph of G versus R as follows:

获取所述R对B相对长分量图如下：Obtain the relative long component graph of R to B as follows:

获取所述R对G相对长分量图如下：Obtain the relative long component graph of R to G as follows:

获取所述B对R相对长分量图如下：Obtain the relative long component graph of B versus R as follows:

获取所述B对G相对长分量图如下：Obtain the relative long component diagram of B to G as follows:

进一步的，所述S34中，判断所述原图P的相对长分量图是否为相对长分量可分图的方法如下：Further, in S34, the method for judging whether the relatively long component graph of the original graph P is a relatively long component separable graph is as follows:

若like

为真，则原图P是相对长分量图，为R对G相对长分量图；此时所述R对G相对长分量图RDCP_R∝G为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of R to G; at this time, the relative long component image RDCP_R∝G of R to G is the mask image of the original image P;

若like

为真，则原图P是相对长分量图，为R对B相对长分量图；此时所述R对B相对长分量图RDCP_R∝B为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of R to B; at this time, the R to B relatively long component image RDCP_R∝B is the mask image of the original image P;

若like

为真，则原图P是相对长分量图，为G对R相对长分量图；此时所述G对R相对长分量图RDCP_G∝R为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of G to R; at this time, the G to R relatively long component image RDCP_G∝R is the mask image of the original image P;

若like

为真，则原图P是相对长分量图，为G对B相对长分量图；此时所述G对B相对长分量图RDCP_G∝B为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of G versus B; at this time, the G versus B relatively long component image RDCP_G∝B is the mask image of the original image P;

若like

为真，则原图P是相对长分量图，为B对G相对长分量图；此时所述B对G相对长分量图RDCP_B∝G为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of B to G; at this time, the relatively long component image RDCP_B∝G of B to G is the mask image of the original image P;

若like

为真，则原图P是相对长分量图，为B对R相对长分量图；此时所述B对R相对长分量图RDCP_B∝R为原图P的掩模图。If it is true, then the original image P is a relatively long component image, which is a relatively long component image of B versus R; at this time, the B versus R relatively long component image RDCP_B∝R is the mask image of the original image P.

有益效果：本发明的一种基于颜色通道运算的去背景方法，利用彩色原图P中分别获取R通道数值r、G通道数值g、B通道数值b，并且充分利用图片的特征，长分量或相对长分量特征，获取原图P的前景，大大地降低了去背景的计算复杂度，能够有效提升用户体验。Beneficial effects: a method for removing background based on color channel calculation of the present invention uses the original color image P to obtain the R channel value r, the G channel value g, and the B channel value b respectively, and makes full use of the characteristics of the picture, long components or Compared with long-component features, obtaining the foreground of the original image P greatly reduces the computational complexity of background removal and can effectively improve user experience.

附图说明Description of drawings

为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained according to these drawings on the premise of not paying creative efforts.

图1为本发明的去背景方法的流程图；Fig. 1 is the flowchart of the background removal method of the present invention;

图2为本发明的实施例中R长分量颜色空间示意图；Fig. 2 is the schematic diagram of R long component color space in the embodiment of the present invention;

图3为本发明的实施例中B长分量颜色空间示意图；Fig. 3 is the schematic diagram of B long component color space in the embodiment of the present invention;

图4为本发明的实施例中G长分量颜色空间示意图；Fig. 4 is a schematic diagram of G long component color space in an embodiment of the present invention;

图5为本发明的实施例中的基本运算单元计算效率比较图；FIG. 5 is a comparison diagram of computing efficiency of basic computing units in an embodiment of the present invention;

图6为本发明的实施例中的去背景实例效果对比图；Fig. 6 is a comparison diagram of the effect of removing the background example in the embodiment of the present invention;

图7为本发明的实施例中的通道运算耗时与距离域运算耗时百分比图；FIG. 7 is a percentage diagram of time-consuming channel calculation and time-consuming distance domain calculation in an embodiment of the present invention;

图8为本发明的实施例中的去背景的流程图。FIG. 8 is a flow chart of background removal in an embodiment of the present invention.

具体实施方式detailed description

为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

基于颜色通道运算法利用通道与通道之间的关系或算术运算来实现背景去除，是一种在精度控制上非常灵活的处理方法，并且不需要指定目标颜色或样本颜色。The color channel-based algorithm uses the relationship between channels or arithmetic operations to achieve background removal. It is a very flexible processing method in terms of precision control, and does not need to specify the target color or sample color.

本实施例提供了一种基于颜色通道运算的去背景方法，如图1和图8所示，包括如下步骤：This embodiment provides a method for removing background based on color channel calculation, as shown in Figure 1 and Figure 8, including the following steps:

S1：获取待处理图片的样本集；并从中选取一张图片作为原图P；S1: Obtain a sample set of pictures to be processed; and select a picture from it as the original picture P;

具体的，本实施例中通过随机取样或者连续取样的方式，形成待处理图片的样本集；Specifically, in this embodiment, a sample set of pictures to be processed is formed by random sampling or continuous sampling;

S2：从所述原图P中分别获取(包含前景和背景的)R通道数值r、G通道数值g、B通道数值b；S2: Obtain the R channel value r, the G channel value g, and the B channel value b (including foreground and background) respectively from the original image P;

S3：根据所述R通道数值r、G通道数值g、B通道数值b；判断是否存在原图P的掩模图；若存在，获取原图P的掩模图，执行S4，否则本发明的基于颜色通道运算的去背景方法不能够使用，执行S6；S3: According to the R channel value r, the G channel value g, and the B channel value b; determine whether there is a mask image of the original image P; if it exists, obtain the mask image of the original image P, and execute S4, otherwise the present invention The background removal method based on color channel calculation cannot be used, go to S6;

具体的为根据原图P的前景背景颜色，来确定通道运算的方法；Specifically, according to the foreground and background colors of the original image P, the method of determining the channel operation;

S4：根据所述原图P的掩模图，获取原图P的理想掩模图；S4: Obtain an ideal mask image of the original image P according to the mask image of the original image P;

具体的，由于所述原图P的掩模图得到的是存在点状噪声的前景掩模，因此对所述原图P的掩模图通过腐蚀操作和膨胀操作去除前景掩模的噪声，以获取理想掩模。其中，腐蚀操作和膨胀操作均为图片处理领域中的常用方法，这里不进行赘述。Specifically, since the mask image of the original image P obtains a foreground mask with point-like noise, the noise of the foreground mask is removed by erosion and expansion operations on the mask image of the original image P, so that Get the ideal mask. Among them, the erosion operation and the dilation operation are common methods in the field of image processing, and will not be repeated here.

S5：将所述原图P的理想掩模图与所述原图P进行掩模运算，获取原图P的前景；执行S6。S5: Perform a mask operation on the ideal mask image of the original image P and the original image P to obtain the foreground of the original image P; execute S6.

具体的，本实施例是利用原图P的理想掩模图与所述原图P进行逻辑“与”运算，得到原图P的前景。Specifically, in this embodiment, the ideal mask image of the original image P is used to perform a logic "AND" operation with the original image P to obtain the foreground of the original image P.

S6：对原图P去背景结束。S6: Remove the background from the original picture P and end.

优选的，所述S3中，获取所述原图P的掩模图的方法如下：Preferably, in the S3, the method for obtaining the mask image of the original image P is as follows:

S31：获取所述原图P的长分量图，包括R长分量图、G长分量图和B长分量图；如图2-4所示；S31: Obtain the long component graph of the original image P, including the R long component graph, the G long component graph, and the B long component graph; as shown in Figure 2-4;

令原图P＝[C_x,y|C_x,y＝(r,g,b)]，其中，C_x,y是像素点；由式(1)Let the original image P=[C_{x, y} |C_{x, y} = (r, g, b)], where C_{x, y} are pixels; by formula (1)

式中，DCP表示长分量图(Dominant Component Picture,DCP)；DCP_R表示R长分量图；DCP_G表示原图P的G长分量图；DCP_B表示B长分量图；C_x,y是像素点的颜色；x为像素点横坐标；y均为像素点的纵坐标；r为R通道数值；g为G通道数值；b为B通道数值；In the formula, DCP means the Dominant Component Picture (DCP); DCP_R means the R long component picture; DCP_G means the G long component picture of the original image P; DCP_B means the B long component picture; C_{x, y} are pixels The color of the point; x is the abscissa of the pixel; y is the ordinate of the pixel; r is the value of the R channel; g is the value of the G channel; b is the value of the B channel;

S32：判断所述原图P的长分量图是否为长分量可分图，若原图P的长分量图是长分量可分图，则存在原图P的掩模图，所述长分量可分图即为原图P的掩模图；否则，执行S33；S32: Determine whether the long component graph of the original image P is a long component separable graph, if the long component graph of the original image P is a long component separable graph, then there is a mask image of the original image P, and the long component can be separated The picture is the mask picture of the original picture P; otherwise, execute S33;

所述S32中，判断原图P的长分量图是否为长分量可分图的方法如下：In said S32, the method for judging whether the long component graph of the original graph P is a long component separable graph is as follows:

若like

为真，则原图P的长分量图为R长分量可分图；此时R长分量图DCP_R为原图P的掩模图；此时使用DCP_R进行去背景；If it is true, the long component graph of the original image P is a separable graph of the R long component; at this time, the R long component graph DCP_R is the mask image of the original image P; at this time, DCP_R is used to remove the background;

式(4)表达的含义如下：对原图P的全部像素点C_x,y进行如下的判断，若该像素C_x,y在前景上，则蕴含有r>g并且r>b关系成立，若该像素不在前景上，即在背景上，则蕴含r<g或者r<b关系成立。由此，可判断是否为R长分量可分图。The meaning expressed by formula (4) is as follows: make the following judgment on all the pixel points C_{x, y} of the original image P, if the pixel C_{x, y} is on the foreground, it implies that r>g and r>b relationship is established, If the pixel is not on the foreground, that is, on the background, it implies that the relationship r<g or r<b is established. From this, it can be judged whether it is an R-long component separable graph.

若like

为真，则原图P的长分量图为G长分量可分图；此时G长分量图DCP_G为原图P的掩模图；此时使用DCP_G进行去背景；If true, the long component graph of the original image P is a G long component separable graph; at this time, the G long component graph DCP_G is the mask image of the original image P; at this time, DCP_G is used to remove the background;

若like

为真，则原图P的长分量图为B长分量可分图；此时B长分量图DCP_B为原图P的掩模图；此时使用DCP_B进行去背景；If it is true, the long component graph of the original image P is a long component separable graph of B; at this time, the long component graph DCP_B of B is the mask image of the original image P; at this time, use DCP_B to remove the background;

式中，

为存在集合中任意一元素；∧为合取运算符；∈为集合的属于运算符；

为蕴含运算符；O为原图P的前景；B为原图P的前背景；In the formula,

is any element in the existence set; ∧ is the conjunction operator; ∈ is the belonging operator of the set;

is the implication operator; O is the foreground of the original image P; B is the foreground and background of the original image P;

在本发明的一个实施例中，如附图6所示前景为手的图片为一张R长分量可分图，理由是手上的像素满足r>g并且r>b，而其他位置像素r<g或者r<b。为了把图6背景去除得到只有前景(手)的图片，即将前景(手)的像素保留，其他像素置为黑色(RGB十进制值为000)，只要将所有像素进行r>g并且r>b的判断，为真的点保留，为假的点置为黑色即可，可以看出，在这个去背景过程中，使用了计算简单的关系运算，避免了使用平方和平方根运算的距离域法，因此，速度快，实时应用场景中用户体验好。In one embodiment of the present invention, as shown in Figure 6, the foreground picture of a hand is an R long component separable map, because the pixels on the hand satisfy r>g and r>b, while other position pixels r <g or r<b. In order to remove the background in Figure 6 to obtain a picture with only the foreground (hand), the pixels of the foreground (hand) are reserved, and the other pixels are set to black (RGB decimal value is 000), as long as all pixels are r>g and r>b Judgment, keep the true points, and set the false points to black. It can be seen that in this background removal process, simple relational operations are used to avoid the distance domain method using square and square root operations. Therefore , high speed, and good user experience in real-time application scenarios.

S33：分别获取原图P的相对长分量图(Relative Dominant Component Picture,RDCP)；所述原图P的相对长分量图包括R对G相对长分量图；R对B相对长分量图；G对R相对长分量图；G对B相对长分量图；B对R相对长分量图；B对G相对长分量图；S33: Acquire the relative long component picture (Relative Dominant Component Picture, RDCP) of the original picture P respectively; the relative long component picture of the original picture P includes R vs. G relative long component picture; R vs. B relative long component picture; G vs. R relative long component graph; G versus B relative long component graph; B versus R relatively long component graph; B versus G relatively long component graph;

获取所述G对B相对长分量图如下：Obtain the relative long component graph of G versus B as follows:

获取所述G对R相对长分量图如下：Obtain the relative long component graph of G versus R as follows:

获取所述R对B相对长分量图如下：Obtain the relative long component graph of R to B as follows:

获取所述R对G相对长分量图如下：Obtain the relative long component graph of R to G as follows:

获取所述B对R相对长分量图如下：Obtain the relative long component graph of B versus R as follows:

获取所述B对G相对长分量图如下：Obtain the relative long component diagram of B to G as follows:

S34：判断所述原图P的相对长分量图是否为相对长分量可分图；若原图P的相对长分量图是相对长分量可分图，则存在原图P的掩模图，所述相对长分量可分图即为原图P的掩模图；若否，则不存在原图P的掩模图。S34: Determine whether the relatively long component graph of the original image P is a relatively long component separable graph; if the relatively long component graph of the original image P is a relatively long component separable graph, then there is a mask graph of the original image P, and the The relatively long component separable graph is the mask graph of the original image P; if not, there is no mask graph of the original image P.

所述S34中，判断所述原图P的相对长分量图是否为相对长分量可分图的方法如下：In said S34, the method for judging whether the relatively long component graph of the original graph P is a relatively long component separable graph is as follows:

若like

为真，则原图P的相对长分量图为R对G相对长分量图；此时所述R对G相对长分量图RDCP_R∝G为原图P的掩模图；此时使用RDCP_R∝G进行去背景；If it is true, then the relative long component graph of the original image P is the relative long component graph of R to G; at this time, the relative long component graph of R to G RDCP_R∝G is the mask image of the original image P; at this time, use RDCP_{R ∝G} to remove the background;

具体的，式(12)表达的含义如下：对原图P的全部像素点C_x,y进行如下的判断，若像素点C_x,y在前景上，则蕴含有r>g关系成立，也就是说，如果像素点r>g，则像素点C_x,y在前景上；若该像素不在前景上，即在背景上，则蕴含r<g关系成立。由此，可判断是否为R对G相对长分量可分图。Specifically, the meaning expressed by formula (12) is as follows: all the pixel points C_{x, y} of the original image P are judged as follows, if the pixel point C_{x, y} is on the foreground, it implies that the relationship r>g is established, and also That is to say, if the pixel point r>g, then the pixel point C_{x, y} is on the foreground; if the pixel is not on the foreground, that is, on the background, then the implication r<g relationship is established. From this, it can be judged whether it is a separable graph with relatively long components of R versus G.

在该类适用例中，为了把背景去除得到只有前景的图片，即将前景的像素保留，其他像素置为黑色(其RGB十进制值为000)，只要将所有像素进行r>g的判断，为真的点保留，为假的点置为黑色即可。同样，可以看出，在这个去背景过程中，使用了计算简单的关系运算，完全避免使用平方和平方根运算，因此，计算复杂度低，速度快，有利于保障实时应用场景中的用户体验。In this type of applicable example, in order to remove the background to obtain a picture with only the foreground, the foreground pixels are reserved, and the other pixels are set to black (the RGB decimal value is 000), as long as all pixels are judged to be r>g, it is true The dots are reserved, and the false dots can be set to black. Similarly, it can be seen that in this background removal process, simple relational operations are used, and square and square root operations are completely avoided. Therefore, the calculation complexity is low and the speed is fast, which is conducive to ensuring user experience in real-time application scenarios.

若like

为真，则原图P的相对长分量图为R对B相对长分量图；此时所述R对B相对长分量图RDCP_R∝B为原图P的掩模图；此时使用RDCP_R∝B进行去背景；If it is true, then the relatively long component graph of the original image P is the relative long component graph of R to B; at this time, the R to B relatively long component graph RDCP_R∝B is the mask image of the original image P; at this time, use RDCP_{R ∝B} to remove the background;

若like

为真，则原图P的相对长分量图为G对R相对长分量图；此时所述G对R相对长分量图RDCP_G∝R为原图P的掩模图；此时使用RDCP_G∝R进行去背景；If it is true, then the relatively long component graph of the original image P is the relative long component graph of G versus R; at this time, the G versus R relatively long component graph RDCP_G∝R is the mask image of the original image P; at this time, use RDCP_{G ∝R} to remove the background;

若like

为真，则原图P的相对长分量图为G对B相对长分量图；此时所述G对B相对长分量图RDCP_G∝B为原图P的掩模图；此时使用RDCP_G∝B进行去背景；If it is true, then the relatively long component graph of the original image P is the relatively long component graph of G versus B; at this time, the G versus B relatively long component graph RDCP_G∝B is the mask image of the original image P; at this time, use RDCP_{G ∝B} to remove the background;

若like

为真，则原图P的相对长分量图为B对G相对长分量图；此时所述B对G相对长分量图RDCP_B∝G为原图P的掩模图；此时使用RDCP_B∝G进行去背景；If it is true, the relatively long component graph of the original image P is the relative long component graph of B to G; at this time, the relative long component graph of B to G RDCP_B∝G is the mask image of the original image P; at this time, use RDCP_{B ∝G} to remove the background;

若like

为真，则原图P的相对长分量图为B对R相对长分量图；此时所述B对R相对长分量图RDCP_B∝R为原图P的掩模图；此时使用RDCP_B∝R进行去背景。If it is true, the relatively long component graph of the original image P is the relative long component graph of B to R; at this time, the relative long component graph of B to R RDCP_B∝R is the mask image of the original image P; at this time, use RDCP_{B ∝R} to remove the background.

基于颜色通道运算去背景方法中，对于原图P，去背景操作后，还包括如下步骤：重复S2～S5，以获取待处理图片样本集中所有样本的前景，在这个过程中，能够对腐蚀操作、膨胀操作等中间过程中的参数进行优化，使得在获取原图P的理想掩模时，能够通过所得到的腐蚀操作、膨胀操作的参数来进行，更进一步的提高获取原图P的理想掩模的速度。In the background removal method based on the color channel calculation, for the original image P, after the background removal operation, the following steps are also included: Repeat S2-S5 to obtain the foreground of all samples in the image sample set to be processed. In this process, the erosion operation can be Optimizing the parameters in the intermediate process such as the erosion operation and the expansion operation, so that when obtaining the ideal mask of the original image P, it can be carried out through the obtained parameters of the erosion operation and the expansion operation, and further improve the ideal mask for obtaining the original image P model speed.

本发明的一个实施例如下：An example of the present invention is as follows:

对颜色通道方案进行评价，使用的测试平台参数为：

Core(TM)i7-9700KCPU,双核，主频均为3.6GHz。To evaluate the color channel scheme, the test platform parameters used are:

Core(TM) i7-9700KCPU, dual-core, the main frequency is 3.6GHz.

使用开发环境：Visual Studio 2017 CommunityUse the development environment: Visual Studio 2017 Community

使用开发语言：C++Use development language: C++

使用相关库：OpenCV 4.5.3Use related library: OpenCV 4.5.3

易用性比较：Ease of use comparison:

表1展示了本实例中的基于颜色通道运算的去背景方法与传统距离域法的易用性对比：Table 1 shows the ease of use comparison between the background removal method based on the color channel operation and the traditional distance domain method in this example:

表1.易用性比较Table 1. Ease of Use Comparison

项目project通道运算法channel algorithm距离域法distance domain method前景色给定难度Foreground color given difficulty简单Simple复杂complex适用场景Applicable scene特定场景specific scene普适Universal计算形式与计算复杂度Calculation Form and Computational Complexity关系运算符/低relational operator/low平方和开平方/高square root/height

其中，通道运算法在选择前景时，是通过颜色通道给定的，相比距离域法，对于原图P存在长分量图和相对长分量图的情况下，本实施例中指定前景非常简单，且计算形式简单。Among them, when the channel algorithm selects the foreground, it is given by the color channel. Compared with the distance domain method, when there are long component maps and relatively long component maps in the original image P, specifying the foreground in this embodiment is very simple. And the calculation form is simple.

本实施例中的基于颜色通道运算的去背景方法，其基本运算单元是关系运算，相比较距离域法的基本运算单元是空间距离公式，如图5所示，由于两者基于的运算方法是不同的，针对不同分辨率的图片进行多次操作后取平均值，得到图中数据是10次基本运算的用时。从图5可以看出，使用基于颜色通道运算的去背景方法的关系运算比距离域法使用的空间距离运算快，优势明显。In the background removal method based on color channel calculation in this embodiment, its basic calculation unit is a relational calculation. Compared with the distance domain method, the basic calculation unit is a spatial distance formula, as shown in Figure 5. Since the calculation methods based on both are The difference is that the average value is taken after performing multiple operations on pictures with different resolutions, and the data in the figure is the time spent on 10 basic operations. It can be seen from Figure 5 that the relational operation using the background removal method based on the color channel operation is faster than the spatial distance operation used by the distance domain method, and has obvious advantages.

表2展示了传统的距离域法与通道运算法在去背景上的效果和时间复杂度的比较。从效果上来说，通道运算法与距离域法效果相当。但是，时间复杂度却相差很多。Table 2 shows the comparison of the effect and time complexity of the traditional distance domain method and the channel algorithm on background removal. In terms of effect, the channel algorithm has the same effect as the distance domain method. However, the time complexity is quite different.

表2.去背景实例对比具体的参数Table 2. Contrast specific parameters for background removal examples

为了验证通道运算法的时间有效性，采用本实施例方法，将120张长分量图和相对长分量图按分辨率进行分组，时间复杂度均值的分布情况如图7所示。由图7可以看出各分辨率下，通道运算耗时仅为距离域运算耗时的4.6％以下。当应用于实时图像识别场景中时，能够有效提升工作效率，用户体验效果非常明显。In order to verify the time effectiveness of the channel algorithm, the method of this embodiment is used to group 120 long-component images and relatively long-component images by resolution, and the distribution of the average time complexity is shown in Figure 7 . It can be seen from Fig. 7 that at each resolution, the channel operation time consumption is only less than 4.6% of the distance domain operation time consumption. When applied to real-time image recognition scenarios, it can effectively improve work efficiency, and the user experience effect is very obvious.

有益效果：本发明基于颜色通道运算的方法通过颜色通道之间的数量关系来反映图片的特征，通过颜色通道的长分量、相对长分量特征来进行颜色空间的筛选，一方面指定目标颜色较为容易，不需要像距离法那样定量采样，另一方面，本发明使用关系运算，计算复杂度大大地降低了，当应用于实时图像识别场景中时，能够有效提升用户体验。Beneficial effects: the method based on the color channel operation of the present invention reflects the characteristics of the picture through the quantitative relationship between the color channels, and screens the color space through the long component and relatively long component characteristics of the color channel. On the one hand, it is easier to specify the target color , does not require quantitative sampling like the distance method. On the other hand, the present invention uses relational operations, which greatly reduces computational complexity. When applied to real-time image recognition scenarios, it can effectively improve user experience.

本文提出的基于颜色通道运算的去背景方法，利用彩色图片像素点(RGB颜色模式)自身的数量关系(即像素点三个颜色分量之间的关系)构造前、背景分离方案。能够充分利用图片的特征，即长分量或相对长分量特征，采用简单的关系运算符实现去背景，显著地提高了去背景的计算复杂度。对由对于背景或前景由相近颜色构成的情况下，去背景场合优势明显，能够很好地优化去背景方法的时间复杂度。在工业控制、医疗服务、人脸识别等新兴应用场景下，对于实时响应、极小误差、良好用户体验等有极致需求时,通过本发明的基于颜色通道运算的方法，能够弥补现阶段传感技术的局限、满足上述场景中大量实时数据的要求。满足“智能化、高性能”这一市场需求。The background removal method based on color channel operation proposed in this paper uses the quantitative relationship of color image pixels (RGB color mode) itself (that is, the relationship between the three color components of pixels) to construct a foreground and background separation scheme. It can make full use of the characteristics of the picture, that is, the long component or relatively long component feature, and use simple relational operators to achieve background removal, which significantly increases the computational complexity of background removal. For the case where the background or foreground is composed of similar colors, the advantage of background removal is obvious, and the time complexity of the background removal method can be well optimized. In emerging application scenarios such as industrial control, medical services, and face recognition, when there is an extreme demand for real-time response, minimal error, and good user experience, the color channel-based calculation method of the present invention can make up for the current sensor Technical limitations, to meet the requirements of a large amount of real-time data in the above scenarios. To meet the market demand of "intelligence, high performance".

最后应说明的是：以上各实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述各实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的范围。Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (6)

Translated fromChinese

1.一种基于颜色通道运算的去背景方法，其特征在于，包括如下步骤：1. a method for removing background based on color channel computing, characterized in that, comprising the steps:S1：获取待处理图片的样本集；并从中选取一张图片作为原图P；S1: Obtain a sample set of pictures to be processed; and select a picture from it as the original picture P;S2：从所述原图P中分别获取R通道数值r、G通道数值g、B通道数值b；S2: Obtain the value r of the R channel, the value g of the G channel, and the value b of the B channel from the original image P;S3：根据所述R通道数值r、G通道数值g、B通道数值b，判断是否存在原图P的掩模图；若存在，获取原图P的掩模图，执行S4，否则执行S6；S3: According to the value r of the R channel, the value g of the G channel, and the value b of the B channel, determine whether there is a mask image of the original image P; if it exists, obtain the mask image of the original image P, and execute S4; otherwise, execute S6;S4：据所述原图P的掩模图，获取原图P的理想掩模图；S4: Obtain an ideal mask image of the original image P according to the mask image of the original image P;S5：将所述原图P的理想掩模图与所述原图P进行掩模运算，获取原图P的前景；执行S6；S5: Perform a mask operation on the ideal mask image of the original image P and the original image P to obtain the foreground of the original image P; execute S6;S6：对原图P去背景结束。S6: Remove the background from the original picture P and end.2.根据权利要求1所述的一种基于颜色通道运算的去背景方法，其特征在于，所述S3中，判断是否存在原图P的掩模图方法如下：2. a kind of method for removing background based on color channel calculation according to claim 1, it is characterized in that, in described S3, judge whether to have the mask method of original picture P as follows:S31：获取所述原图P的长分量图，包括R长分量图、G长分量图和B长分量图；S31: Obtain the long component graph of the original image P, including the R long component graph, the G long component graph, and the B long component graph;S32：判断所述原图P的长分量图是否为长分量可分图，若原图P的长分量图是长分量可分图，则存在原图P的掩模图，所述长分量可分图即为原图P的掩模图；否则，执行S33；S32: Determine whether the long component graph of the original image P is a long component separable graph, if the long component graph of the original image P is a long component separable graph, then there is a mask image of the original image P, and the long component can be separated The picture is the mask picture of the original picture P; otherwise, execute S33;S33：获取原图P的相对长分量图；S33: Obtain a relatively long component image of the original image P;S34：判断所述原图P的相对长分量图是否为相对长分量可分图；若原图P的相对长分量图是相对长分量可分图，则存在原图P的掩模图，所述相对长分量可分图即为原图P的掩模图；若否，则不存在原图P的掩模图。S34: Determine whether the relatively long component graph of the original image P is a relatively long component separable graph; if the relatively long component graph of the original image P is a relatively long component separable graph, then there is a mask graph of the original image P, and the The relatively long component separable graph is the mask graph of the original image P; if not, there is no mask graph of the original image P.3.根据权利要求2所述的一种基于颜色通道运算的去背景方法，其特征在于，所述S31中，获取所述原图P的长分量图如下：3. a kind of background removal method based on color channel calculation according to claim 2, is characterized in that, in described S31, obtains the long component figure of described original picture P as follows:令原图P＝[C_x,y|C_x,y＝(r,g,b)]，其中，C_x,y是原图P的像素点；则Let the original image P=[C_{x, y} |C_{x, y} = (r, g, b)], where C_{x, y} are the pixels of the original image P; then

式中，DCP_R表示R长分量图；DCP_G表示原图P的G长分量图；DCP_B表示原图P的B长分量图；C_x,y是原图P的像素点；x为像素点横坐标；y均为像素点的纵坐标；r为R通道数值；g为G通道数值；b为B通道数值。In the formula, DCP_R represents the R long component graph; DCP_G represents the G long component graph of the original image P; DCP_B represents the B long component graph of the original image P; C_{x, y} is the pixel point of the original image P; x is the pixel The abscissa of the point; y is the ordinate of the pixel; r is the value of the R channel; g is the value of the G channel; b is the value of the B channel.4.根据权利要求3所述的一种基于颜色通道运算的去背景方法，其特征在于，所述S32中，判断原图P的长分量图是否为长分量可分图的方法如下：4. A kind of method for removing background based on color channel calculation according to claim 3, characterized in that, in said S32, the method for judging whether the long component graph of the original image P is a long component separable graph is as follows:若like

为真，则原图P的长分量图为R长分量可分图；If is true, the long component graph of the original graph P is a long component separable graph of R;若like

为真，则原图P的长分量图为G长分量可分图；is true, the long component graph of the original graph P is a long component separable graph of G;若like

为真，则原图P的长分量图为B长分量可分图；is true, the long component graph of the original graph P is a separable graph of B long component;式中，

为存在集合中任意一元素；∧为合取运算符；∈为集合的属于运算符；

为蕴含运算符；O为原图P的前景；B为原图P的前背景。In the formula,

is any element in the existence set; ∧ is the conjunction operator; ∈ is the belonging operator of the set;

is the implication operator; O is the foreground of the original image P; B is the foreground and background of the original image P.5.根据权利要求4所述的一种基于颜色通道运算的去背景方法，其特征在于，所述S33中，获取原图P的相对长分量图如下：5. a kind of background removal method based on color channel operation according to claim 4, is characterized in that, in described S33, obtains the relatively long component figure of original picture P as follows:所述原图P的相对长分量图包括R对G相对长分量图、R对B相对长分量图、G对R相对长分量图、G对B相对长分量图、B对R相对长分量图、B对G相对长分量图；The relatively long component graphs of the original image P include R to G relatively long component graphs, R to B relatively long component graphs, G to R relatively long component graphs, G to B relatively long component graphs, and B to R relatively long component graphs , B to G relative long component graph;获取所述G对B相对长分量图如下：Obtain the relative long component graph of G versus B as follows:

获取所述G对R相对长分量图如下：Obtain the relative long component graph of G versus R as follows:

获取所述R对B相对长分量图如下：Obtain the relative long component graph of R to B as follows:

获取所述R对G相对长分量图如下：Obtain the relative long component graph of R to G as follows:

获取所述B对R相对长分量图如下：Obtain the relative long component graph of B versus R as follows:

获取所述B对G相对长分量图如下：Obtain the relative long component diagram of B to G as follows:

6.根据权利要求5所述的一种基于颜色通道运算的去背景方法，其特征在于，所述S34中，判断所述原图P的相对长分量图是否为相对长分量可分图的方法如下：6. A method for removing background based on color channel calculation according to claim 5, characterized in that in said S34, it is a method of judging whether the relatively long component graph of the original image P is a relatively long component separable graph as follows:若like

为真，则原图P是相对长分量图，为R对G相对长分量图；此时所述R对G相对长分量图RDCP_R∝G为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of R to G; at this time, the relative long component image RDCP_R∝G of R to G is the mask image of the original image P;若like

为真，则原图P是相对长分量图，为R对B相对长分量图；此时所述R对B相对长分量图RDCP_R∝B为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of R to B; at this time, the R to B relatively long component image RDCP_R∝B is the mask image of the original image P;若like

为真，则原图P是相对长分量图，为G对R相对长分量图；此时所述G对R相对长分量图RDCP_G∝R为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of G to R; at this time, the G to R relatively long component image RDCP_G∝R is the mask image of the original image P;若like

为真，则原图P是相对长分量图，为G对B相对长分量图；此时所述G对B相对长分量图RDCP_G∝B为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of G versus B; at this time, the G versus B relatively long component image RDCP_G∝B is the mask image of the original image P;若like

为真，则原图P是相对长分量图，为B对G相对长分量图；此时所述B对G相对长分量图RDCP_B∝G为原图P的掩模图；If it is true, then the original image P is a relatively long component image, which is a relatively long component image of B to G; at this time, the relatively long component image RDCP_B∝G of B to G is the mask image of the original image P;若like

为真，则原图P是相对长分量图，为B对R相对长分量图；此时所述B对R相对长分量图RDCP_B∝R为原图P的掩模图。If it is true, then the original image P is a relatively long component image, which is a relatively long component image of B versus R; at this time, the B versus R relatively long component image RDCP_B∝R is the mask image of the original image P.

Images