Image impulse noise self-adaptive filtering method based on genetic algorithm-back propagation (GA-BP) neural network

Abstract

The invention discloses an image impulse noise self-adaptive filtering method based on a genetic algorithm-back propagation (GA-BP) neural network. The method includes a first step of optimizing an initial weight value and an initial threshold value of a BP neural network through a genetic algorithm, a second step of training the CA-BP neural network, a third step of conducting noise detection on image through the well trained GA-BP neural network, and a fourth step of conducting self-adaptive average filtering on image noise points which are detected out. Due to the fact that the initial weight value and the initial threshold value of the BP neural network are poetized through the genetic algorithm, the problems of local minimum and slow speed convergence and the like are solved through a BP algorithm, and generalization ability of the neural network can be effectively improved.

Description

technical field [0001] The invention relates to an image noise filtering method, more specifically an image salt and pepper noise self-adaptive filtering method based on GA-BP neural network. Background technique [0002] Images are often affected by salt and pepper noise in the process of acquisition and transmission. It is a basic problem in image processing to effectively remove the noise in the image while maintaining its characteristics. Median filtering is one of the most commonly used methods to remove salt and pepper noise in images. The operation is simple and efficient, but the filtering effect is affected by the size of the filtering window, and median filtering processes all pixels uniformly, and both signal points and noise points are changed. , resulting in the blurring of the edge details of the image. In view of this defect of median filtering, the idea of ​​noise location first and then filtering came into being. Based on this, scholars have proposed a var...

AI Technical Summary

Problems solved by technology

The filtering quality of these algorithms and the ability to protect image details have been improved, but they all have their own limitations in practical applications, especially when the noise density is large, the filtering effect is not ideal.

[0003] BP neural network has strong generalization ability, simple structure and easy implementation, and is suitable for solving the classification problem of judging whether an image pixel is a noise point or a signal point, but the optimization of BP neural network is not unique, and there will be local minimum phenomenon; on the other hand , the initial weights and thresholds are randomly given, resulting in more training times and slow convergence

Images