Texture Extraction and Analysis by Statistical Methods for Fish Species Classification

For the classification of fish species, textures were extracted and analyzed in this paper. Images were obtained by wireless GPRS model while users sending their photographs to our system. In addition, all the images were pre-processed by dividing them into different rectangle texture images. Than Statistical texture of fish are extracted and analyzed to describe the different fish species in images through higher-order moments of their grayscale histograms (GH) and gray level co-occurrence matrices (GLCM). The experiment result indicate that, for those images with little gray distribution, histogram equalization should be done to increase the range of the gray level before image compression. It is no matter that how the users chose a shooting angle and the feature data should be normalized as inputs for a robust classifier.