Improved LBP used for texture feature extraction
Liu Ha · Computer Engineering and Applications Journal · 2014
For most texture feature extraction method, the problem of high computational complexity always exists. An Improved Complete Local Binary Pattern algorithm(ICLBP)is proposed based on Complete LBP(CLBP). ICLBP preserves the CLBP_S in CLBP, while makes an improvement on CLBP_M and CLBP_C, and proposes a new texture description operator ICLBP_T. ICLBP can describe the local texture feature in a comprehensive way, and the problem that CLBP_M operator in CLBP is sensitive to uneven distribution of gray, is well solved in ICLBP. The classification results on Outex and CURet image databases suggest that, compared to the existing LBP algorithm, ICLBP has obtained a higher classification accuracy, meanwhile, the ICLBP SCT feature in ICLBP has a lower feature dimension and better practical value.