Study On Texture Recognition Based On Co-occurrence Extension Eight Neighborhood Matrix

Texture exture recognition is a basic research direction in machine vision and image processing. As a very effective and successful method, it has been applied in several areas, especially in biometric feature detection, remote sensing and image processing. Intelligent information technology has been developing rapidly in recent years. Therefore, image recognition with computers instead of the human has been increasingly concerned. As one research content of image processing and intelligent recognition, texture feature extraction has an important research prospect and wide application value.In this paper, we introduce four algorithms of texture image processing and texture image classification algorithm based on the characteristics of image texture firstly. We mainly focus on the statistic method of the image texture analysis and feature extraction method based on eight- neighbor.Considering the rotational invariance of texture feature recognition, we propose a feature extraction method based on the co-occurrence extension of eight neighborhood matrix. Based on the gray co-occurrence matrix, this method compute its extension of eight-neighbor matrix and then do SVM and K-neighbor classification.We compare our method with several classic texture feature extraction methods on the CURe T, K TH-TIPS and UIUC texture data sets. Both supervised and unsupervised recognition method have been adopted. The results of the experiments show that our proposed method achieves good results in the classification of texture image. When the percentage of the training set and test set decreases, our method is also robust. In the last part of this thesis, we apply our method to face recognition on the ORL, AR, FERET face data sets. And the recognition results are good.