The Extraction And Application Of Image Texture Feature Based On Lbp

As one of the fundamental property of the object surface, texture feature reflects the structural organization of the surface arrangement about the significant information and the relationship with its surroundings. The extraction of texture feature is describing the image texture quantitatively or qualitatively through some techniques about image processing. As one of the classical texture analysis method proposed in 1990s, Local Binary Pattern (LBP) is nonparametric descriptor which is non-sensitive to noise and is calculated quickly. Based on these, LBP has been widely applied to texture-based images processing in recent years.In this thesis, the theory of LBP is applied to fuzzy classification about criminal investigation images and retrieval about texture image, and the main points are as follows:1. The main research of this paper is about the extraction and application of image texture feature based on LBP. With the condition of having studied and analyzed about the features of criminal investigation images and the advantages of fuzzy classification, LBP and 2-level wavelet are fused to extract the texture features and fuzzy N classifier is used to classify the criminal investigation images scene to get the fuzzy uncertainty of the images at last, preliminarily reaching the aim of automatic management of the criminal investigation images.2. Because LBP will discard the float relationship in the neighbor sets while presents the texture just using the gray value differences among the central pixel and its neighbors, which results in various gray distributions of central pixel correspond to the same LBP code. In most cases, this phenomenon will lead to specific images texture discarding, information wasting and worse performance. Therefore, the comparion between objects when characterizing local texture not only relies on the central pixel, but also merges the floating-point relationship between neighboring pixels. The new proposed Float-LBP extracts more useful texture information while reserves the better capabilities of LBP, and the simulation results prove that the modified method makes the retrieval performance better.