Research On Image Retrieval Algorithms Combining Directional Enhanced LBP And Gabor Wavelet Features

This paper studies and discusses an image retrieval algorithm based on texture features.In order to achieve efficient extraction of texture features,a combination of spatial and frequency domain features is selected to describe the texture features.After analyzing the advantages and disadvantages of various texture feature extraction methods,the direction-enhanced local binary pattern(Local Binary Pattern,LBP)and Gabor wavelet transform features are selected as the description algorithm of texture features.As a classical algorithm in spatial domain texture feature extraction,LBP algorithm has the advantages of gray invariance and fast calculation speed,but it has the disadvantages of anti-noise ability,flexibility,poor stability and lack of direction recognition.After studying the algorithm mode,the analysis of various improved algorithms is summarized,and the improvement direction of the algorithm is summarized.A more stable direction-enhanced LBP algorithm is selected as an important algorithm for texture feature extraction.The algorithm not only includes the comparison of the gray value of the center point and the neighborhood point,but also the gray value comparison between the neighborhood points.It can improve the anti-noise ability of the algorithm based on the recognition of the enhanced direction.As a classical algorithm for signal frequency domain processing,Gabor wavelet transform has strong stability and denoising ability,and can effectively multi-scale and multi-directional transform images.However,as a single feature extraction algorithm,there is a problem that it is easy to ignore the local texture features of the image.The combination of direction-enhanced LBP and Gabor wavelet transform can realize the complementarity of the two algorithms.It increases the multi-directional and multi-scale global feature extraction while increasing the extraction of local detail features,and improves the quality of texture description from multiple angles.By using the Corel-1k database to compare the combined algorithms used in this paper,the precision is used as a measure of the performance of the algorithm.It is verified by experiments that under the same conditions,the image retrieval algorithm used in this paper has a good image retrieval effect.