Image Segmentation based on Genetic Algorithm


One of the most difficult tasks in image processing is the determination of a suitable set of features which can be used to segment images. In this research, the standard deviation that represents feature of image has been used in image segmentation as indicator to determine the isolation of one object from another or an object from a background. This feature has been used by a Genetic Algorithm (GA) to become a fitness function that will help in searching process for the optimal solution. The value of standard deviation is high in the case of a difference between various diverse regions of the image and small in one region. Using this feature in maximizes the difference among different regions and minimizes the interclass variance, a GA is used to evolve a sub-image convolution kernel to produce kernel with a best features that can be used in the segmentation of image. The space-filling curve approach has been used to convert the kernel from a one dimensional (1-D) form into a two dimensional (2-D) form. The evolution process of a genetic algorithms are done on a kernel in array form, while the convolution process between a kernel and image is used a kernel in a matrix form.