Many techniques in pattern recognition

robot vision

segmentation

feature extraction and etc require edge detection as a basic instrument. Although many methods have been suggested

the performance is quite different for different types of images and there is still not a general method. In this paper

we proposed a novel edge processing approach which makes the detected edge maps more valid and more ideal

instead of introducing a new edge detection method. The proposed method uses genetic algorithm to optimize the edge maps after edge detection. First

it encodes the edge maps into a two|dimensional binary array and determines the fitness based on valid edge structural templates for each individual. Second

the parent population is generated by changing a small part of pixels in edge maps randomly. Then the proposed method re|allocates edge points according to the genetic operators such as crossover and mutation

and forms their offspring population. Finally

elitist section is adopted to drive the genetic procedure approaching convergent state. When the genetic algorithm is converged

the optimized edge maps can be obtained and the noises in edge maps can be effectively reduced. The proposed method has been carried out for both the artificial and natural images

and the experimental results have shown its good performance.