| Color image segmentation technology based on texture is always the focus of academicresearch, many scholars have put forward a series of algorithms, divided into thesegmentation algorithm based on statistical texture, texture segmentation algorithm based onmodel and method based on signal processing. The texture segmentation algorithm based onstatistics is the most important, the most basic algorithm, in a certain sense is the basis of theother two algorithms, the statistical texture segmentation algorithm based on imagesegmentation algorithm, Markov random chain theory to the segmentation efficiency isrelatively high, relatively high cutting precision algorithm. Compared with the segmentationalgorithm of image in classical Markov chains, multiscale Markov chain algorithm for imagesegmentation of multi-scale theory was introduced into the algorithm, to solve the traditionalMarkov chain method can only show the segmentation results in single scale problem, butalso in the multi-scale segmentation for all complete segmentation results are based on pixel,texture regions the relationship between the optimized segmentation results at all levels, thesegmentation accuracy has been improved greatly.Multiscale Markov chain algorithm for image segmentation using the relationshipbetween the single pixel color information based on Markov chain to complete the form forpreliminary segmentation of the most important step, because the algorithm only uses pixelcolor information in the texture fragmentation process to complete clustering, making thelack of space distribution of the gray. Information about the spatial distribution information ofgray, color spatial distribution is used to describe the texture is not stable, different regions ofthe same texture image is easily affected by illumination angle and shooting angle resulting ina color change. Therefore, the algorithm of texture segmentation results shows overlysensitive to color, prone to erroneous segmentation.Texture is the essence characteristics of the object are different from each other inimages, still lacks of a clear definition for the texture. However, the texture can be seen as acombination of complex visual entity or pattern, combination is similar images. According tothe description, the clustering characteristic of TFR algorithm fragmentation phase lacks ofspatial relations between based on gray scale, texture element (Texture-Primitive) multilayerMarkov chain texture segmentation method. The color and the gray distribution feature spaceis constructed by combining the texture primitives, as the texture of the object fragments clustering. Gray distribution feature primitives texture pattern and contrast LBP/C valuesfrom the local two (Local Binary Pattern and Contrast) expression. For texture elementsegmentation enables fragmentation results stable, provides correct basis and further fragmentprocessing to construct8direction for regional subsequent Markov chain. Experiments provethat the algorithm proposed in this paper, the segmentation accuracy are improved obviously.Because the gray distribution information of texture used in foundation treatment phasealgorithm, overcomes the error segmentation texture caused by factors such as light, forremote sensing image segmentation, the segmentation results of stability algorithms performbetter. |
