Image Segmentation Method for Non-destructive Testing Based on Pixel Block Clustering Incorporating Class Index Suppression Factor

Abstract

The invention discloses a pixel block clustering non-destructive detection image segmentation method incorporating an exponential suppression factor. The original grayscale image is input, and for any pixel block, the spatial position relationship and grayscale relationship between pixels in the block are used to obtain the pixels in the block. Weight, traverse all pixel blocks, construct the class index inhibitory factor expression that is inversely related to the class size, integrate the class index inhibitory factor, each pixel block and its weight reconstruction into the objective function of the fuzzy mean, establish a new objective function and obtain According to the new membership degree, the image is segmented based on iterative fuzzy mean clustering, and the result is output. The present invention simultaneously considers the characteristics of large differences between the background and the target area to construct an index-like suppression factor, integrates the pixel block and the suppression factor into the objective function of the fuzzy mean to perform image segmentation, is robust to noise, and can be effectively applied to image segmentation for non-destructive testing . The invention is applicable to the technical field of identifying and measuring targets in images.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a pixel block clustering nondestructive detection image segmentation method incorporating an exponential suppression factor. Identify and measure targets. Background technique [0002] Non-destructive testing images generally refer to images formed by the use of ray, eddy current, ultrasound, infrared and other technologies to measure and identify defects, objects, and other instruments, materials, and equipment. Segmenting this type of image is a necessary prerequisite for automatic target measurement and recognition, but the imaging environment is very complex and changeable during the imaging process, and is easily affected by a variety of unfavorable factors, such as poor working conditions, dim light, and poor equipment. Thermal radiation, etc., so the obtained non-destructive testing images are often accompanied by noise interference. In addition, the...

AI Technical Summary

Problems solved by technology

This kind of method inevitably introduces one or more adjustable parameters because of the consideration of neighborhood information, which will cause damage to the adaptability of the algorithm.

[0004] The existing fuzzy C-means method treats each pixel "equally", so that the large-area background pixels in the non-destructive testing image have far more influence on the objective function than the small-area target pixels, which will make the clustering center of the target pixels easier Offset due to background pixels, resulting in ineffective segmentation

Images