Research And Applpication On Texture Extraction Algorithm Based On Self-adaptive Compression Of Grayscales

As an important visual clue, texture feature is difficult to describe, but common existsin images. It is regularly reflected in the distribution of pixel gray scale and theneighborhood. Texture analysis is a lively research direction in image processing, machinevision, image analysis and retrieval areas. Through researches in recent decades, it hasmade great progress, and many texture feature extraction methods have born. However,there are more or less some problems in variety of texture analysis methods,and it is stilldifficult to extract texture information from the image rapidly and accurately. This paperfocuses on how to improve the effectiveness of texture feature extraction. The maincontents are as follows.Firstly, this paper analyzes the process of extracting image texture feature usingGLCM(Gray Level Co-occurrence Matrix) which is one of the statistical methods,combining with the characteristics of spectral methods, improves traditional grayscalecompression algorithm to achieve a multi-resolution texture feature extraction, proposes anew adaptive grayscale compression algorithm based on distribution of grayscale toimprove the discrimination and effectiveness in the process of grayscale compression.Secondly, combining with GLCM, this paper processes the compressed gray imageand presents a texture feature extraction algorithm based on the adaptive grayscalecompression. The algorithm constructs a GLCM in view of the grayscale image andextracts quadratic statistical parameters from the matrix as the texture feature vector,containing both local texture details and overall texture of the image, and obtaining textureinformation on different scales. In the context of Morton code based on quadtree, it alsocan be used to divide an image into multi-level partitions with the merging standards ofthe similarity of texture feature vector, in order to represent the picture information withless data.Finally, two experiments are designed to verified the theory proposed. One is theimage retrieval experiment which makes a comprehensive analysis of the precision incompare with traditional texture feature extraction algorithm, and it is projected to testifythe accuracy of the proposed algorithm in extracting texture feature. In the other experiment the new algorithm is applied to the grayscale compression process of a singleimage by quadtree coding.