技术领域technical field
本发明涉及图像处理领域，更具体地，涉及一种人脸识别方法。The present invention relates to the field of image processing, and more particularly, to a face recognition method.
背景技术Background technique
生物识别技术是通过人类生物特征识别身份的一种技术；人脸识别技术涉及图像处理、计算机技术、心理学和生物学等领域，是当前模式识别和人工智能领域的重要研究课题；人脸与人体的其它生物特征(指纹、虹膜、掌纹、视网膜等)一样与生俱来，它的唯一性和不易被复制的良好特性为身份鉴别提供了必要的前提；目前，人脸识别广泛应用于国家安全、合法监控、金融、公安等领域；但是人脸识别依然存在较大困难，具体体现在：Biometric technology is a technology that identifies identity through human biometrics; face recognition technology involves image processing, computer technology, psychology and biology and other fields, and is an important research topic in the current pattern recognition and artificial intelligence fields; Other biological features of the human body (fingerprint, iris, palm print, retina, etc.) are innate, and its uniqueness and good characteristics that are not easy to be copied provide the necessary premise for identification; at present, face recognition is widely used in National security, legal surveillance, finance, public security and other fields; however, face recognition still has great difficulties, which are embodied in:
(1)人脸特征稳定性较差；(1) The stability of facial features is poor;
人脸表面为柔性皮肤，可塑性强，因此当人们的表情、化妆、年龄等条件改变时，人脸也会随之改变，这给人脸识别很大的难度。The surface of the human face is flexible skin with strong plasticity, so when people's expressions, makeup, age and other conditions change, the human face will also change, which makes face recognition very difficult.
(2)可靠性低、安全性较低；(2) Low reliability and low security;
虽说人脸是人体独一无二的特征，每个人的面部特征各不相同，但从表面上来看人类的脸都有相似的结构，全球人口基数大，使得许多人脸差异较小，这给人脸识别的可靠性、安全性带来了极大的挑战。Although the human face is a unique feature of the human body, and each person's facial features are different, on the surface, human faces have a similar structure. The global population base is large, making many faces less different, which gives face recognition. reliability and security have brought great challenges.
(3)受外界条件影响大；(3) It is greatly affected by external conditions;
人脸识别技术应用时的环境因素如光线条件、遮挡物、环境等都会严重影响人脸识别的效果。Environmental factors such as light conditions, occlusions, and the environment during the application of face recognition technology will seriously affect the effect of face recognition.
LBP(Local Binary Pattern)即局部二值模式是一种基于纹理特征的人脸识别方法；最初是由Ojala等人提出局部二值式(LBP)并将其用于纹理分类，LBP的显著特点是对光照具有不变性；后来被ahonend等人应用于光照人脸识别中，并取得了一定效果；LBP已经成功应用于人脸识别，唇语识别，表情检测，动态纹理等等领域；而LBP-HF(局部二值模式傅里叶直方图)算法是基于LBP方法的一种衍生品，其在保证计算速度的同时，比LBP具有更高的识别率。LBP (Local Binary Pattern) is a face recognition method based on texture features; it was originally proposed by Ojala et al. and used for texture classification. The salient features of LBP are It is invariant to illumination; it was later applied to illumination face recognition by ahonend and others, and achieved certain results; LBP has been successfully applied to face recognition, lip recognition, expression detection, dynamic texture and other fields; and LBP- The HF (Local Binary Mode Fourier Histogram) algorithm is a derivative based on the LBP method, which has a higher recognition rate than LBP while ensuring the calculation speed.
光照归一化是利用基本的图像处理技术对图像处理进行预处理，获取鲁棒的光照图像；是由Meylan等人提出的一种基于的视网膜模型的局部对比度增强方法；该方法首先利用小波变换获取人脸图像的低频子带和高频子带，然后对低频子带进行对比度增强处理，对高频子带进行阈值截断处理，最后将处理后的低频子带和高频子带进行二维离散小波反变换，从而重建出人脸图像。假设人脸图像经过二维小波变换得到近似子带A、水平子带DH、垂直子带DV和对角子带DD四个子带；其中A为低频子带，其他三个为高频子带。Illumination normalization is to use basic image processing technology to preprocess the image to obtain a robust illumination image; it is a local contrast enhancement method based on retinal model proposed by Meylan et al. The method first uses wavelet transform Obtain the low-frequency sub-band and high-frequency sub-band of the face image, then perform contrast enhancement processing on the low-frequency sub-band, perform threshold truncation processing on the high-frequency sub-band, and finally perform two-dimensional processing on the processed low-frequency sub-band and high-frequency sub-band. Inverse discrete wavelet transform to reconstruct the face image. It is assumed that the face image is subjected to two-dimensional wavelet transform to obtain four sub-bands: approximate sub-band A, horizontal sub-band DH, vertical sub-band DV and diagonal sub-band DD; where A is the low-frequency sub-band, and the other three are high-frequency sub-bands.
其中，人脸识别的光照问题严重影响了人脸识别系统的性能。Among them, the lighting problem of face recognition seriously affects the performance of the face recognition system.
发明内容SUMMARY OF THE INVENTION
本发明为克服上述现有技术所述的人脸识别的光照问题严重影响了人脸识别系统的性能的缺陷，提供一种人脸识别方法。The present invention provides a face recognition method in order to overcome the defect of the above-mentioned prior art that the illumination problem of face recognition seriously affects the performance of the face recognition system.
所述方法包括以下步骤：The method includes the following steps:
S1：将原始人脸图像灰度化；S1: Grayscale the original face image;
S2：对经S1灰度化后的图像，进行光照归一化，得到新的灰度人脸图像；S2: Perform illumination normalization on the grayscaled image of S1 to obtain a new grayscale face image;
S3：求取新的灰度人脸图像的LBP-HF(局部二值模式傅里叶直方图)特征值；并对新的灰度人脸图像实现基于GPU加速的Gabor滤波，得到空间域结果；S3: Obtain the LBP-HF (Local Binary Mode Fourier Histogram) eigenvalues of the new grayscale face image; implement GPU-accelerated Gabor filtering for the new grayscale face image, and obtain the spatial domain result ;
S4：结合整个人脸图像的LBP-HF特征值和空间域结果进行人脸识别。S4: Combine the LBP-HF eigenvalues of the entire face image and the spatial domain results for face recognition.
优选地，S2包括以下步骤：Preferably, S2 includes the following steps:
S2.1：对低频子带进行处理，包括：感光层处理、外网状层处理、边缘增强；对高频子带进行阈值截断处理；S2.1: Process the low-frequency sub-band, including: photosensitive layer processing, outer mesh layer processing, and edge enhancement; perform threshold truncation processing on the high-frequency sub-band;
S2.2：对处理过后的低频子带和高频子带使用二维离散小波反变换重建人脸图像；S2.2: Reconstruct the face image using two-dimensional discrete wavelet inverse transform on the processed low-frequency sub-bands and high-frequency sub-bands;
S2.3：对重建后的人脸图像进行几何归一化，得到新的灰度人脸图像。S2.3: Perform geometric normalization on the reconstructed face image to obtain a new grayscale face image.
优选地，S2.1包括以下步骤：Preferably, S2.1 includes the following steps:
S2.1.1：对低频子带进行感光处理，处理公式为：S2.1.1: Perform photosensitive processing on the low frequency sub-band, the processing formula is:
其中，Amax(x,y)是A(x,y)中的最大值，表示A(x,y)中某点的调节因子，ρ1是高斯分布的标准差系数，Ap(x,y)是对应的输出值，A(x,y)是低频子带位置(x,y)的灰度值，低频子带对应图像的轮廓；where Amax (x,y) is the maximum value in A(x,y), Represents the adjustment factor of a point in A(x, y), ρ1 is the standard deviation coefficient of the Gaussian distribution, Ap (x, y) is the corresponding output value, A(x, y) is the low-frequency subband position (x , y) gray value, the low-frequency subband corresponds to the contour of the image;
S2.1.2：以S2.1.1中的Ap(x,y)作为输入，对低频子带进行外网状层处理，处理公式为：S2.1.2: Take Ap (x, y) in S2.1.1 as the input, and perform outer mesh layer processing on the low frequency subband. The processing formula is:
其中，是Ap(x,y)上的最大值，表示Ap(x,y)上某个调节因子，ρ2是高斯分布标准系数，Ao(x,y)是对应输出值；in, is the maximum value on Ap (x,y), Represents a certain adjustment factor on Ap (x, y), ρ2 is the standard coefficient of Gaussian distribution, and Ao (x, y) is the corresponding output value;
S2.1.3：边缘增强：对S2.1.2中的输出值Ao(x,y)进行高斯差分滤波：S2.1.3: Edge enhancement: Gaussian difference filtering on the output value Ao (x,y) in S2.1.2:
其中ρL、ρH分别是对应于两个高斯分布标准差系数，A'(x,y)为处理过后的低频子带。Among them, ρL and ρH are the standard deviation coefficients corresponding to two Gaussian distributions, respectively, and A'(x, y) is the processed low-frequency subband.
优选地，S2.2的具体过程为：首先分别对水平子带DH、垂直子带DV和对角子带DD三个高频子带进行阈值截断，公式为：Preferably, the specific process of S2.2 is as follows: first, threshold truncation is performed on the three high-frequency sub-bands of the horizontal sub-band DH, the vertical sub-band DV and the diagonal sub-band DD respectively, and the formula is:
t＝median(|H(x,y)|)t=median(|H(x,y)|)
其中median(.)是返回矩阵的中间值，DH'、DV'、DD'是水平子带、垂直子带和对角子带处理后的高频子带；H'(x,y)为截断后的高频子带的频率，H(x,y)为截断前的高频子带的频率，t为阈值；Where median(.) is the middle value of the returned matrix, DH', DV', DD' are the high-frequency sub-bands processed by the horizontal sub-band, the vertical sub-band and the diagonal sub-band; H'(x, y) is the truncated sub-band The frequency of the high frequency subband of , H(x, y) is the frequency of the high frequency subband before truncation, and t is the threshold;
然后，对处理过后的低频子带A'和高频子带DH'、DV'、DD'使用二维离散小波反变换重建出的人脸图像。Then, the processed low-frequency sub-band A' and the high-frequency sub-bands DH', DV', DD' are used to reconstruct the face image using the two-dimensional discrete wavelet inverse transform.
优选地，S3中新的灰度人脸图像的LBP-HF特征值的求取过程为：Preferably, the process of obtaining the LBP-HF feature value of the new grayscale face image in S3 is as follows:
S3.1.1：对新的灰度人脸图像分块；S3.1.1: Block the new grayscale face image;
S3.1.2：通过局部保持投影算法来提取各个子区域的LBP-HF特征值；S3.1.2: Extract the LBP-HF eigenvalues of each sub-region through a local-preserving projection algorithm;
S3.1.3：将得到的各LBP-HF的特征值按照顺序排列连接成一个一维向量，从而得到整个人脸图像的LBP-HF特征值。S3.1.3: Arrange and connect the obtained eigenvalues of each LBP-HF in order to form a one-dimensional vector, thereby obtaining the LBP-HF eigenvalues of the entire face image.
优选地，S3中对新的灰度人脸图像实现基于GPU加速的Gabor滤波，得到空间域结果；的具体过程为：Preferably, in S3, GPU-accelerated Gabor filtering is implemented for the new grayscale face image to obtain a spatial domain result; the specific process is:
S3.2.1：对新的人脸图像和Gabor核基于GPU的FFT转换，得到目标图像；S3.2.1: GPU-based FFT transformation of the new face image and Gabor kernel to obtain the target image;
S3.2.2：将空间域的卷积操作转化为频率域的相乘操作，得到频域结果；S3.2.2: Convert the convolution operation in the space domain into a multiplication operation in the frequency domain to obtain the frequency domain result;
S3.2.3：将得到的频域结果基于GPU的FFT逆转换，得到空间域结果。S3.2.3: Inversely transform the obtained frequency domain result based on the GPU's FFT to obtain the spatial domain result.
优选地，在执行S3.2.1-S3.2.3的操作的同时使用GPU加速并行处理技术，对FFT变换进行加速，同时将Gabor核的FFT变换结果保存在显存中，减少运行时间。Preferably, while performing the operations of S3.2.1-S3.2.3, the GPU accelerated parallel processing technology is used to accelerate the FFT transformation, and at the same time, the FFT transformation result of the Gabor core is stored in the video memory, thereby reducing the running time.
与现有技术相比，本发明技术方案的有益效果是：本发明所述方法基于光照归一化和GPU加速的LBP-HF特征实现对人脸的识别，解决各种光照变化对人脸识别的影响和改善人脸识别速率的问题；灰度化原人脸图像，对灰度人脸图像进行光照归一化处理，得到新的灰度人脸图像，对新的灰度人脸图像分块并提取各个子区域的LBP-HF特征值，并将得到的LBP-HF特征值按照顺序排列，从而得到整个人脸图像的LBP-HF特征值，与此同时对新的灰度人脸图像进行GPU加速操作得到空间域结果，最后结合起来进行人脸识别；本发明能够很大程度上消除各种光照变化对人脸识别的影响，并提高人脸识别速率。Compared with the prior art, the beneficial effects of the technical solution of the present invention are: the method of the present invention realizes the recognition of the face based on the LBP-HF feature of illumination normalization and GPU acceleration, and solves the problem of face recognition caused by various illumination changes. The influence of the grayscale face image and the problem of improving the face recognition rate; grayscale the original face image, perform illumination normalization processing on the grayscale face image, and obtain a new grayscale face image. block and extract the LBP-HF eigenvalues of each sub-region, and arrange the obtained LBP-HF eigenvalues in order, so as to obtain the LBP-HF eigenvalues of the entire face image. The GPU acceleration operation is performed to obtain the spatial domain result, and finally the face recognition is combined; the present invention can largely eliminate the influence of various illumination changes on the face recognition, and improve the face recognition rate.
附图说明Description of drawings
图1为本发明所述人脸识别方法的流程图。FIG. 1 is a flow chart of the face recognition method according to the present invention.
具体实施方式Detailed ways
附图仅用于示例性说明，不能理解为对本专利的限制；The accompanying drawings are for illustrative purposes only, and should not be construed as limitations on this patent;
为了更好说明本实施例，附图某些部件会有省略、放大或缩小，并不代表实际产品的尺寸；In order to better illustrate this embodiment, some parts of the drawings are omitted, enlarged or reduced, which do not represent the size of the actual product;
对于本领域技术人员来说，附图中某些公知结构及其说明可能省略是可以理解的。It will be understood by those skilled in the art that some well-known structures and their descriptions may be omitted from the drawings.
下面结合附图和实施例对本发明的技术方案做进一步的说明。The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.
本实施例提供一种人脸识别方法，如图1所示，所述方法包括以下步骤：This embodiment provides a face recognition method, as shown in FIG. 1 , the method includes the following steps:
S1：将原始人脸图像灰度化；S1: Grayscale the original face image;
S2：光照归一化；S2: Light normalization;
1.对低频子带的处理；1. Processing of low frequency subbands;
(1)感光层处理：低频子带的处理公式如下：(1) Photosensitive layer processing: The processing formula of the low frequency sub-band is as follows:
其中，Amax(x,y)是A(x,y)中的最大值，表示A(x,y)中某点的调节因子，ρ1是高斯分布的标准差系数，Ap(x,y)是对应的输出值，A(x,y)是低频子带位置(x,y)的灰度值，低频子带对应图像的轮廓；where Amax (x,y) is the maximum value in A(x,y), Represents the adjustment factor of a point in A(x, y), ρ1 is the standard deviation coefficient of the Gaussian distribution, Ap (x, y) is the corresponding output value, A(x, y) is the low-frequency subband position (x , y) gray value, the low-frequency subband corresponds to the contour of the image;
(2)外网状层处理：上一步的Ap(x,y)作为输入：(2) Outer mesh layer processing: Ap (x, y) of the previous step is used as input:
其中，是Ap(x,y)上的最大值，表示Ap(x,y)上某个调节因子，ρ2是高斯分布标准系数，Ao(x,y)是对应输出值。in, is the maximum value on Ap (x,y), Represents a certain adjustment factor on Ap (x, y), ρ2 is the standard coefficient of Gaussian distribution, and Ao (x, y) is the corresponding output value.
(3)边缘增强：对上一步的输出值Ao(x,y)进行高斯差分滤波：(3) Edge enhancement: Gaussian difference filtering is performed on the output value Ao (x, y) of the previous step:
其中ρL、ρH分别是对应于两个高斯分布标准差系数，A'(x,y)为处理过后的低频子带。Among them, ρL and ρH are the standard deviation coefficients corresponding to two Gaussian distributions, respectively, and A'(x, y) is the processed low-frequency subband.
2.对高频子带的处理；2. Processing of high frequency sub-bands;
分别对各个高频子带进行阈值截断，具体的定义如下：Threshold truncation is performed on each high-frequency subband respectively, and the specific definitions are as follows:
t＝median(|H(x,y)|)t=median(|H(x,y)|)
其中median(.)是返回矩阵的中间值，DH'、DV‘、DD'是相应处理后的高频子带。where median(.) is the middle value of the returned matrix, and DH', DV', DD' are the corresponding processed high frequency subbands.
最后对处理过后的低频子带A'和高频子带DH'、DV‘、DD'使用二维离散小波反变换重建人脸图像。Finally, the processed low-frequency sub-band A' and high-frequency sub-bands DH', DV', DD' are used to reconstruct the face image by using two-dimensional discrete wavelet inverse transform.
3.将人脸图像几何归一化；3. Normalize the face image geometry;
通过基于双三次插值的方法，将重建的人脸图像转换为150*150分辨率，作为S3中提取LBP-HF特征值和实现基于GPU加速的Gabor滤波所用。Through the method based on bicubic interpolation, the reconstructed face image is converted to 150*150 resolution, which is used for extracting LBP-HF eigenvalues in S3 and implementing GPU-accelerated Gabor filtering.
S3：对S2中归一化后的人脸图像进行划分，切割成相同大小的子区域；对人脸图像所切割出的子区域，提取它们的LBP-HF(局部二值模式傅里叶直方图)特征值；将人脸图像各个子区域的LBP-HF特征值按顺序连接成一个一维向量，用于人脸识别；S3: Divide the normalized face image in S2 and cut it into sub-regions of the same size; extract their LBP-HF (Local Binary Mode Fourier Histogram) for the sub-regions cut out from the face image Figure) eigenvalues; the LBP-HF eigenvalues of each sub-region of the face image are sequentially connected into a one-dimensional vector for face recognition;
同时，对S2中归一化后的人脸图像实现基于GPU加速的Gabor滤波；At the same time, GPU-accelerated Gabor filtering is implemented for the normalized face image in S2;
Gabor滤波需要大量分别对Gabor核实部和虚部的卷积操作，Gabor核越大，卷积图像越大，耗时就越长，本实施例选用21*21像素大小的Gabor核，卷积图像为150*150。针对卷积操作时间长的特点，使用基于FFT的方法，将空间域的卷积操作转化为频率域的相乘操作，每次卷积只需要一次FFT变换，一次相乘和一次逆次FFT变换，时间复杂度为nlog(n)，这样能达到较快的速度，同时使用GPU加速并行处理技术，对FFT变换进行加速，同时将Gabor核的FFT变换结果保存在显存中，减少运行时间。Gabor filtering requires a large number of convolution operations on the Gabor verification part and the imaginary part respectively. The larger the Gabor kernel, the larger the convolution image and the longer the time. is 150*150. In view of the long time of convolution operation, the method based on FFT is used to convert the convolution operation in the spatial domain into a multiplication operation in the frequency domain. Each convolution only needs one FFT transformation, one multiplication and one inverse FFT transformation. , the time complexity is nlog(n), which can achieve faster speed. At the same time, the GPU accelerated parallel processing technology is used to accelerate the FFT transformation, and at the same time, the FFT transformation result of the Gabor core is saved in the video memory to reduce the running time.
基于GPU加速的Gabor滤波就是为了加速得到空域结果，并行加快速度，为了达到最好的并行优化效果，最好的情况是针对每一个像素点的操作进行并行化，线程之间不需要交互，没有数据上的相关性。The Gabor filtering based on GPU acceleration is to accelerate the spatial results and speed up the speed in parallel. In order to achieve the best parallel optimization effect, the best case is to parallelize the operation of each pixel, and there is no need for interaction between threads. correlation in the data.
S4：结合整个人脸图像的LBP-HF特征值和空间域结果进行人脸识别。S4: Combine the LBP-HF eigenvalues of the entire face image and the spatial domain results for face recognition.
附图中描述位置关系的用语仅用于示例性说明，不能理解为对本专利的限制；The terms describing the positional relationship in the accompanying drawings are only used for exemplary illustration, and should not be construed as a limitation on this patent;
显然，本发明的上述实施例仅仅是为清楚地说明本发明所作的举例，而并非是对本发明的实施方式的限定。对于所属领域的普通技术人员来说，在上述说明的基础上还可以做出其它不同形式的变化或变动。这里无需也无法对所有的实施方式予以穷举。凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明权利要求的保护范围之内。Obviously, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the embodiments of the present invention. For those of ordinary skill in the art, changes or modifications in other different forms can also be made on the basis of the above description. There is no need and cannot be exhaustive of all implementations here. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910244760.XACN110046559A (en) | 2019-03-28 | 2019-03-28 | A kind of face identification method |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910244760.XACN110046559A (en) | 2019-03-28 | 2019-03-28 | A kind of face identification method |
| Publication Number | Publication Date |
|---|---|
| CN110046559Atrue CN110046559A (en) | 2019-07-23 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910244760.XAPendingCN110046559A (en) | 2019-03-28 | 2019-03-28 | A kind of face identification method |
| Country | Link |
|---|---|
| CN (1) | CN110046559A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348783A (en)* | 2020-10-27 | 2021-02-09 | 基建通(三亚)国际科技有限公司 | Image-based person identification method and device and computer-readable storage medium |
| CN115439938A (en)* | 2022-09-09 | 2022-12-06 | 湖南智警公共安全技术研究院有限公司 | Anti-splitting face archive data merging processing method and system |
| CN117253279A (en)* | 2023-11-17 | 2023-12-19 | 蓝色火焰科技成都有限公司 | Vehicle remote control method, device and medium based on vehicle-mounted information service unit |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1790374A (en)* | 2004-12-14 | 2006-06-21 | 中国科学院计算技术研究所 | Face recognition method based on template matching |
| WO2011055164A1 (en)* | 2009-11-06 | 2011-05-12 | Vesalis | Method for illumination normalization on a digital image for performing face recognition |
| CN103207986A (en)* | 2013-02-06 | 2013-07-17 | 江苏清大维森科技有限责任公司 | Face recognition method based on local binary pattern-histogram Fourier (LBP-HF) |
| CN103870820A (en)* | 2014-04-04 | 2014-06-18 | 南京工程学院 | Illumination normalization method for extreme illumination face recognition |
| CN104881660A (en)* | 2015-06-17 | 2015-09-02 | 吉林纪元时空动漫游戏科技股份有限公司 | Facial expression recognition and interaction method based on GPU acceleration |
| WO2016110005A1 (en)* | 2015-01-07 | 2016-07-14 | 深圳市唯特视科技有限公司 | Gray level and depth information based multi-layer fusion multi-modal face recognition device and method |
| CN107239732A (en)* | 2017-04-18 | 2017-10-10 | 广东工业大学 | A kind of tired expression recognition method based on Gabor characteristic and rarefaction representation |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1790374A (en)* | 2004-12-14 | 2006-06-21 | 中国科学院计算技术研究所 | Face recognition method based on template matching |
| WO2011055164A1 (en)* | 2009-11-06 | 2011-05-12 | Vesalis | Method for illumination normalization on a digital image for performing face recognition |
| CN103207986A (en)* | 2013-02-06 | 2013-07-17 | 江苏清大维森科技有限责任公司 | Face recognition method based on local binary pattern-histogram Fourier (LBP-HF) |
| CN103870820A (en)* | 2014-04-04 | 2014-06-18 | 南京工程学院 | Illumination normalization method for extreme illumination face recognition |
| WO2016110005A1 (en)* | 2015-01-07 | 2016-07-14 | 深圳市唯特视科技有限公司 | Gray level and depth information based multi-layer fusion multi-modal face recognition device and method |
| CN104881660A (en)* | 2015-06-17 | 2015-09-02 | 吉林纪元时空动漫游戏科技股份有限公司 | Facial expression recognition and interaction method based on GPU acceleration |
| CN107239732A (en)* | 2017-04-18 | 2017-10-10 | 广东工业大学 | A kind of tired expression recognition method based on Gabor characteristic and rarefaction representation |
| Title |
|---|
| 周巍等: "基于光照归一化分块完备LBP特征的人脸识别", 《计算机工程与应用》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348783A (en)* | 2020-10-27 | 2021-02-09 | 基建通(三亚)国际科技有限公司 | Image-based person identification method and device and computer-readable storage medium |
| CN112348783B (en)* | 2020-10-27 | 2022-08-05 | 基建通(三亚)国际科技有限公司 | Image-based person identification method and device and computer-readable storage medium |
| CN115439938A (en)* | 2022-09-09 | 2022-12-06 | 湖南智警公共安全技术研究院有限公司 | Anti-splitting face archive data merging processing method and system |
| CN115439938B (en)* | 2022-09-09 | 2023-09-19 | 湖南智警公共安全技术研究院有限公司 | Anti-splitting face archive data merging processing method and system |
| CN117253279A (en)* | 2023-11-17 | 2023-12-19 | 蓝色火焰科技成都有限公司 | Vehicle remote control method, device and medium based on vehicle-mounted information service unit |
| Publication | Publication Date | Title |
|---|---|---|
| Mahmood et al. | Facial expression recognition in image sequences using 1D transform and gabor wavelet transform | |
| Xie et al. | Normalization of face illumination based on large-and small-scale features | |
| Jain et al. | Natural image denoising with convolutional networks | |
| Du et al. | Adaptive region-based image enhancement method for robust face recognition under variable illumination conditions | |
| Zaied et al. | A novel approach for face recognition based on fast learning algorithm and wavelet network theory | |
| Liang et al. | A novel framework based on wavelet transform and principal component for face recognition under varying illumination | |
| Cheng et al. | Robust face recognition based on illumination invariant in nonsubsampled contourlet transform domain | |
| CN110046559A (en) | A kind of face identification method | |
| CN106204482A (en) | Based on the mixed noise minimizing technology that weighting is sparse | |
| CN110674824A (en) | Finger vein segmentation method and device based on R2U-Net and storage medium | |
| CN105631441A (en) | Human face recognition method | |
| Sulistianingsih et al. | Classification of batik image using grey level co-occurrence matrix feature extraction and correlation based feature selection | |
| Kar et al. | Facial expression recognition system based on variational mode decomposition and whale optimized KELM | |
| KR102629393B1 (en) | Parallel subsampling structured cnn based finger-vein recognition method | |
| Hu et al. | Multi-level feature fusion facial expression recognition network | |
| Wang et al. | Face Recognition Method Based on Improved Gabor Wavelet Transform Algorithm. | |
| CN102214292B (en) | Illumination processing method for human face images | |
| Sun et al. | Facial expression recognition based on histogram sequence of local Gabor binary patterns | |
| CN111950454A (en) | A Finger Vein Recognition Method Based on Bidirectional Feature Extraction | |
| CN101231695A (en) | Iris identification method based on multi-resolution analysis | |
| Al-Dabbas et al. | High-accuracy models for iris recognition with merging features | |
| Zhao et al. | A Wavelet-Based Image Preprocessing Method for Illumination Insensitive Face Recognition. | |
| Jin et al. | Illumination robust face recognition using random projection and sparse representation | |
| Cheng et al. | Illumination normalization for face recognition under extreme lighting conditions | |
| CN111950320A (en) | An Iris Recognition Method Based on Relative Total Variation and Probabilistic Co-Representation |
| 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 | Application publication date:20190723 | |
| RJ01 | Rejection of invention patent application after publication |