Liu Tao: a Novel Feature Extraction Algorithm for Haar Local Binary Pattern Texture

The locality and edges of texture image may be ignored by the Haar local binary pattern texture features owing to strong subjectivity and poor ability to self-adaptive to the artificial setting judgment threshold. Therefore, from the perspective of Human Vision System (HVS), the new Haar local binary pattern texture feature extraction algorithm (HLBP_HVS) is proposed. The local and global structure information of images are obtained and the self-adaptive and local optimal judgment threshold are calculated by the analysis of HVS influence factors, which include: texture detail and distribution of spatial position. The new Haar local binary pattern texture feature HLBP_HVS, which is objective and conforms to the image texture details and distribution of spatial position, could be extracted. The experimental results show that the proposed algorithm can effectively avoid the influence of the artificial judgment threshold on the texture detail and reflects the structure information of the image. Through comparison and analysis of the test results, we suggest that the accuracy of classification for Brodatz texture library also can be further improved.