Robust Local Binary Pattern Algorithms For Texture Image Classification

Texture is existed widely,and the surface of most objects can be regarded as texture,which directly reflects properties of the objects.In recent years,as one of texture features,the local binary pattern(LBP)has been widely applied to pattern recognition fields.However,LBP is still not robust enough in some specific conditions.For example,the spatial information is still not fully considered,and the anti-noise ability is not strong enough.Therefore,this article deeply studies the LBP by adding the global information and enhancing the anti-noise ability.As the LBP only analyzes the local feature of image and ignores the spatial distribution information of image,it is difficult to describe the image comprehensively.Therefore,this paper applies the idea of combining local and global feature for reference,and improves the representation ability of feature by combining the local binary pattern with the global feature.Based on the combining of the global gray-level histogram with the LBP,a feature named global and local binary pattern(GLBP)is proposed.The GLBP not only contains the local information of image but also includes the spatial distribution information.The performance of GLBP is evaluated on the Outex texture database.In addition,in order to enhance the discriminant ability against the noise,we propose an efficient texture feature named noise-tolerant complete enhanced local binary pattern(CELBPNT).CELBPNT is robust to illumination,rotation and noise.Its feature extraction process involves the following three steps.First,different patterns in LBP are reclassified to form an enhanced LBP(ELBP)based on their structures and frequencies.Then,in order to describe the local feature completely and sufficiently,the difference of modulus value and the center pixel information are added to ELBP to develop a complete ELBP feature,named CELBP.Meanwhile,the adaptive threshold of CELBP is determined by the image size.Finally,CELBPNT is proposed by using the favorable characteristics of multi-scale analysis on CELBP.The features are evaluated on the popular Outex and CUReT databases with different intensity and different types of noises.Extensive experimental results show that the CELBPNT not only demonstrates superior performance under no-noise condition but also effectively improves the performance in present of noise due to its high robustness and distinctiveness.Finally,our work is summarized,and future work is also analyzed and discussed.