Multi-threshold image segmentation method based on crossover mutation artificial fish swarm algorithm

Abstract

The invention discloses a multi-threshold image segmentation method based on a crossover mutation artificial fish swarm algorithm and mainly aims to solve the problem that in the prior art, information loss of a segmented image is serious. The method comprises the implementation steps that 1, an image is input, and pixel grayscale values at all image pixel points are acquired; 2, c thresholds are selected to segment the image into c+1 classes; 3, n artificial fishes are generated, and each artificial fish is a 1xc-dimension vector and represents a group of threshold possible solutions; 4, a fitness function made according to the kapur maximum entropy criterion is regarded as a goal, and a maximum value of the fitness function is searched for; and 5, a group of thresholds corresponding to the fitness maximum value found through search are utilized to perform image segmentation, the pixel points with the grayscale values in the same interval are classified into one class, and the segmented image is output. Through the method, the optimizing precision of the artificial fish swarm algorithm in the optimizing process is effectively improved, the image segmentation effect is improved further in combination with multi-threshold image segmentation, and the method can be applied to computer vision analysis.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a method for segmenting multi-threshold images, which can be used in computer vision analysis. Background technique [0002] Multi-threshold image segmentation technology is to divide the image into several meaningful areas to ensure the effective follow-up work. It is an important step from image processing to image analysis. Among them, the main idea of ​​the multi-threshold image segmentation method is to select multiple appropriate thresholds, and attribute the pixels whose gray value is between the two thresholds to the same class, so that the selected threshold after segmentation can satisfy the maximum entropy of kapur or satisfy The formula of the Otsu criterion, and then map the classified results back to the space of the original image, thus obtaining the final image segmentation result. [0003] The traditional two-dimensional threshold image segmentati...

AI Technical Summary

Problems solved by technology

[0006] Therefore, more and more researchers combine the swarm intelligence algorithm with multi-threshold image segmentation. For example, Horng proposed in the paper "Multilevel thresholding selection based on the artificial bee colony algorithm for image segmentation." Expert Systems with Applications 38.11 (2011) By combining the artificial bee colony algorithm with the idea of ​​multi-threshold image segmentation, the artificial bee colony algorithm has strong exploration ability, but weak development ability, so the optimization efficiency in the iterative process is poor

In "GrayscaleImage Segmentation Using Multilevel Thresholding and Nature-Inspired Algorithms."Hybrid Soft Computing for Image Segmentation. SpringerInternational Publishing, 2016.23-52. In "GrayscaleImage Segmentation Using Multilevel Thresholding and Nature-InspiredAlgorithms. The gravity search hybrid algorithm has strong development ability, but weak exploration ability, so the optimization efficiency is only strong in the early stage of the iterative process, and it is difficult to maintain a long-term high optimization efficiency

According to the kapur maximum entropy criterion, the multi-threshold image segmentation is realized. Compared with the traditional exhaustive method, the time required for the calculation process of the multi-threshold image segmentation algorithm based on the swarm intelligence algorithm is greatly shortened, but because the multi-threshold image segmentation algorithm based on the swarm intelligence algorithm Segmentation itself is difficult to balance the exploration ability and development ability in the optimization process, resulting in the problem that it is impossible to find the best possible solution for the optimal threshold value, resulting in serious loss of image information after the obtained threshold value segmentation, and the segmentation effect is not good enough. Further impacting the process of computer vision analysis

Images