CLBP texture image processing method based on multi-scale thresholds

Abstract

The invention discloses a CLBP texture image processing method based on multi-scale thresholds. The method comprises the following steps: in a training stage and a testing stage, a KNN classifier is trained in the training stage; in the test stage, similar operation is carried out on an input test texture image data set, after features are extracted through CLBP operation based on multi-scale threshold values, the features are input into a trained KNN classifier, and an image label of a test texture image is output. According to the method provided by the embodiment of the invention, the extracted texture features contain the local region and global information of the texture image at the same time through the CLBP method based on the multi-scale threshold, and the texture information extracted by the features is more comprehensive, so that the classification accuracy is improved.

Description

technical field [0001] The invention relates to texture image classification technology, in particular to a multi-scale threshold-based CLBP texture image processing method. Background technique [0002] Texture is an inherent feature of almost all natural surfaces and contains important information about the surface structure of an image. As a basic problem in image processing, pattern recognition, computer recognition and other related fields, texture classification has a very wide range of applications in the fields of fabric detection, content-based image retrieval, remote sensing and medical image analysis. [0003] Real-world texture images may be taken in various lighting environments and pose angles, therefore, a good texture classification method usually needs to be able to handle grayscale, rotation and scale variations. Since the 1960s, researchers have proposed a variety of texture classification methods to solve these problems, which can be roughly divided into...

AI Technical Summary

Problems solved by technology

[0005] However, in the traditional CLBP, the threshold in the sign encoding scheme of local difference is a single point pixel value, which only describes the pixel-level difference of the image, and the obtained features are too local, and lack the expression of image region or global information. In addition, the threshold scheme for single-point pixels is susceptible to interference from image transformations, resulting in poor feature stability; while the other two components use global thresholds, which lose feature details and regional information.

Therefore, the threshold scheme of CLBP is not fully descriptive of image features, so its classification results are not accurate

Images