Some of the material in is restricted to members of the community. By logging in, you may be able to gain additional access to certain collections or items. If you have questions about access or logging in, please use the form on the Contact Page.
In order to better study how to improve the accuracy of texture recognition in image processing, we will study the concept of information entropy, which characterizes the uncertainty of the information source. Since the information entropy value of different images and local parts of an image are different, we wonder whether the information entropy will contain some features suitable for distinguishing the different textures of images. If so, we can improve the accuracy of texture recognition by using the information entropy. When we mention texture, people will think of the wood grain on wooden furniture immediately, or the decorative pattern on flowery cloth, and so on. Wood grain is a natural texture, while the decorative pattern is an artificial texture. They both reflect some changes of the surface color and grayscale throughout the image. These changes are characteristics of the image itself. The texture is very important, but difficult part of image processing. The recognition and selection of texture features is always a difficult problem to quantity in the field of image processing. In recent years, with the texture in high resolution remote sensing images, medical image detection, as well as in the textile industry, it has an increasingly wide range of applications. Therefore, it is very critical to study the directions in texture recognition research. In this thesis, we design an experiment to obtain the effect of entropy on texture recognition. The main texture analysis method used in this experiment is the gray level co-occurrence matrix (GLCM) method. This is a method to describe texture by analyzing the spatial correlation characteristics of grayscale images and to recognize the texture by these features. In this thesis, we will select 5 original images with different textures. We generate 1000 different sub-images after image segmentation as training and test samples from these original images. By constructing the GLCM matrix for each sub-image, texture features of each sub-image can be obtained, and then appropriate texture features can be selected through feature extraction and feature selection. Then, half of these samples were added with entropy as a new texture feature, and the other half were not added with entropy. Finally, we determine the accuracy of these two types of classification using a standard classifier. Thus, we can obtain the effect of entropy in texture recognition based on the GLCM method.