Abstract: Aiming at the problems of low tumor recognition rate and slow convergence speed caused by limited samples of intestinal tumor images, we proposed a feature extraction method of intestinal tumor images based on local binary pattern (LBP) and gray level co-occurrence matrix (GLCM). Firstly, the maximum interclass variance method was used to automatically calculate the image gray threshold and extract the region of interest. Secondly, the LBP+GLCM was used to extract the features of partial intestinal tumor image, and support vector machine was used for recognition. Experimental results of  1 500 intestinal tumor images show that this method can achieve 94.84% recognition accuracy and can effectively assist medical diagnosis and treatment.

Key words: intestinal tumor, region of interest, local binary patterns (LBP), grey level co-occurrence matrix (GLCM), support vector machines (SVM)