I am not giving you an answer, instead, I want to know more about the objective function. What and how are you measuring? I think this information will help you get better answers. Or is there a standard objective when talking about images? 

Best,

The answer might be that the treatment of images - such as the famous restoration/recovery problems - are better done through the use of nonlinear optimization techniques. Some row-action methods have even been famously shown in Hollywood movies - see, e.g., No Way Out with Harrison Ford, in which a picture of him is restored at Pentagon by a row-action method. I would definitely NOT begin by trying to fit a meta-heuristic, when there are viable alternatives.

Gaussian noise is added to images then the objective is using genetic algorithm (crossover and mutation operators) to decrease the noise. I am using MSE(Minimum squared error) to compare them.

Michael, thank you for the suggestion. But the idea was to explore image enhancement from genetic algorithm perspective.

There are many opinions about what you said about the crossover and mutation ( global search and local search), some researchers with what you said, and the other have a different perspective. So:

Any effective search algorithm trying to find the global peaks must

use two techniques: Exploration and Exploitation. Effective algorithms are

concerned with finding a balance between these two concepts.

1. Exploration: gaining access to new and promising areas in the search

space. The goal of this technique is diversification in order to avoid falling

into premature convergence. This is a global search.

2. Exploitation: using knowledge that has been gained from exploration to

find new best points. This process uses the term “refining” (intensification)

to investigate a specific region; in other words, this is a local search.

A study by Back (1994) showed that the selection method (selection

pressure) controls the degree of exploitation and exploration together,

where increasing or decreasing the selection pressure leads to exploitation

or exploration respectively.

(Eiben & Schippers, 1998) discussed how to achieve these two

concepts, and it was observed that the common opinion was that crossover

and mutation operations are used for exploration, while selection is used

for exploitation. From their point of view, this matter (to achieve

exploration and exploitation) needs a deep understanding of evolutionary

algorithms.

The previous two works of research and other research that has been

mentioned in the research paper (Črepinšek, Liu, & Mernik, 2013 ". Exploration and

exploitation in evolutionary algorithms: a survey  "), where

the aim of this paper is to provide treatment for exploration and

exploitation, discussed several issues, including: how to achieve

exploration and exploitation in Evolutionary Algorithms (EAs)? How to

control exploration and exploitation? and achieving exploration and

exploitation balance through diversity.

To learn more about new types of crossover and mutation ,  see  these papers:

https://www.researchgate.net/publication/307598841_Enhancing_Genetic_Algorithms_using_Multi_Mutations_Experimental_Results_on_the_Travelling_Salesman_Problem

https://www.researchgate.net/publication/310465457_On_Enhancing_Genetic_Algorithms_Using_New_Crossovers

Article Enhancing Genetic Algorithms using Multi Mutations: Experime...

Article On Enhancing Genetic Algorithms Using New Crossovers

Thank you.

After all, I used Genetic Algorithms with PSNR to set fitness function.