Gabor Filters and Grey-level Co-occurrence Matrices in Texture Classification

Texture classification is a problem that has been studied and tested using different methods due to its valuable usage in various pattern recognition problems, such as wood recognition and rock classification. The Grey-level Co-occurrence Matrices (GLCM) and Gabor filters are both popular techniques used on texture classification. This paper combines both techniques in order to increase the accuracy. The paper used 32 textures from the Brodatz texture dataset with 1024 training samples and 1024 testing samples. GLCM achieved a recognition rate of 84.00%, Gabor filters achieved 79.58% while combination of GLCM and Gabor filters achieved a recognition rate of 88.52%, which is better than both methods. The experiments showed that the best result can be achieved by using a GLCM with grey level of 16, spatial distance of one pixel and combine with Gabor features decomposed to six features.