Research On Image Feature Extraction Method Based On Local Information

Image feature extraction is an important part of image processing and the basis for solving a series of computer vision tasks such as image classification and recognition.Nowadays,the continuous development of image acquisition technology makes more and more images collected under complex conditions,and people pay more and more attention to the influence of interference factors such as lighting changes and occlusions on image feature extraction.Compared with the global features of the image,the local features of the image can not only better represent the detail information,but also obtain the real image features affected by various interference factors well.Both Local Binary Pattern(LBP)and Gabor wavelet transform have certain local information description capabilities,and are widely used in the field of feature extraction.This paper mainly used LBP and Gabor wavelet transform as the basis to study the image feature extraction method under the interference of lighting,occlusion,attitude and other changing factors.The specific work is as follows:First,the basic principles of feature extraction methods based on LBP and Gabor wavelet transform are introduced in this paper.In view of their problems in feature extraction,several improved methods are introduced.In the face recognition field,these methods are simulated experimentally,the advantages and disadvantages of different methods are analyzed,and the theoretical foundation is laid for subsequent research.Secondly,in order to solve the problem of poor performance of feature extraction methods caused by single interference such as lighting or occlusion,this paper proposes a Gabor wavelet transform image feature extraction method based on Sub-images Correlation Analysis(SCA-Gabor).This method uses the idea of CS-LBP to analyze the sub-images obtained by Gabor wavelet transform.It combines the advantages of CS-LBP and Gabor wavelet transform to effectively retain the information between Gabor wavelet sub-images.This method uses the strategy of overlapping blocks and histograms to obtain the final feature vector,which reduces the feature vector dimension while extracting rich local information.Simulation experiments in the field of face recognition prove the effectiveness of the method.Finally,under the influence of multiple interference factors such as light and occlusion,this paper combines the two-dimensional discrete wavelet transform(2D-DWT)and integral projection.For different image databases,using a fixed block size is easy to cause local information loss,the SCA-Gabor image feature extraction based on varying scales(VSSCA-Gabor)is proposed.This method uses the physical meaning of 2D-DWT subbands and combines the idea of integral projection to extract the information on the low frequency subbands to more specifically block the image.The experiment proves that the proposed VSSCA-Gabor method can get higher recognition rate in the face databases.