Gray level fluctuation threshold segmentation method of image with non-uniform illumination

Abstract

The invention provides a gray level fluctuation threshold segmentation method of an image with non-uniform illumination. Compared with the existing local algorithm, the method regards the image as a three dimensional terrain of the gray level, the non-uniform illumination is equivalent to changing of the terrain landform, the grey value fluctuates up or down within a local area, thus a larger-scale crest point and trough in each gray scale fluctuation curve are searched along the horizontal and vertical directions in an iterative way, a reasonable threshold between each pair of crest point and trough point is calculated to realize the division of image goal pixels and background pixels, and finally, threshold images obtained in the two directions are subjected to integration to obtain a final segmentation image. By applying the method, the influence of non-uniform illumination can be effectively reduced, the instantaneity of the algorithm is improved, and an effective pretreatment technology base for target detection of the image under a complex background is provided.

Description

Technical field [0001] The invention belongs to the field of image segmentation in machine vision inspection, and specifically relates to a gray-scale fluctuation threshold used to process images of uneven industrial illumination in image analysis systems (such as target defect detection systems, character recognition systems, target positioning measurement systems) Segmentation method. Background technique [0002] Image threshold segmentation is an important preprocessing part of machine vision inspection technology, and it plays a vital role in the subsequent processing accuracy. Among them, the global algorithm uses information such as the variance or entropy of the entire image to delimit the category of the pixel, while the local algorithm determines the category of the pixel through the local information of the image. Compared with the global algorithm, the local algorithm has obvious advantages in processing the uneven illumination detection images taken in the complex i...

AI Technical Summary

Problems solved by technology

This local algorithm is not only sensitive to parameter adjustment, but also has poor real-time performance.

Professor Huang of the Chinese Academy of Sciences adaptively adjusts the block size based on the pyramid structure of the Lorentz information degree, and uses the Otsu algorithm to divide each sub-block. This method reduces the influence of uneven illumination to a certain extent, but still cannot eliminate discontinuity

Dr. Chou of Tamkang University in Taiwan used support vector machine to determine sub-block categories and adopted different segmentation strategies for sub-blocks, but this method requires a lot of prior knowledge and is not very practical

Images