Research On Algorithms For Image Segmentation Based On Renyi Theory

Image segmentation is a crucial and important part in image processing and computer vision. It is a challenging and classic task in artificial intelligence field. The fundamental aim of image segmentation is dividing the target area and background area to extracting the part what we interested. Image segmentation could make images better to be understood and analyzed. After several decades of development, there have been lots of image segmentation algorithms. Among these algorithms, people pay much attention on thresholding methods which is simple and practical.This paper is main focusing on the subject closely: ―The research on algorithms for image segmentation based on Renyi theory‖. In order to simplify the computational complexity and improve the segmentation effect, this paper has obtained two corresponding improved algorithms. The effectiveness of each algorithm has been proved by relevant experiments. This paper’s main work can be listed as follows:① To select the parameter in decomposed 2D Renyi entropy image thresholding segmentation method, a new adaptive method is proposed using particle swarm optimization algorithm, according to the uniformity measure which is an image segmentation evaluation criteria. The experiment results show that the method not only can get the suitable parameter α and desired segmentation result for each image but also can reduce the computational complexity from6O(L) to 2O(L).And the computational time of this method is only one-ten-thousandth of the computational time of 2D Renyi entropy image thresholding segmentation method with adaptive selection parameter.② Aiming at the limitation of one-dimensional Renyi entropy algorithm the paper researched a new iterative method in image segmentation based on one-dimensional Renyi’s threshold. The method iteratively searches for sub regions of the image for segmentation for processing to get the final segmentation threshold. The iterative method starts with Renyi’s threshold and computes the grey values with the greatest frequency of the two classes as separated by the threshold. Based on the Renyi’s threshold and the two values with the greatest frequency of the two classes, the method separates the image into three classes. They are background, object and to-be-determined region. The first two classes will not be processed further. The to-be-determined region will be processed at next iteration. At next iteration, Renyi’s method is applied on the to-be-determined region to calculate a new threshold and grey values with the greatest frequency of two new classes. And the to-be-determined region is again separated into three classes, namely, background, object and a new to-be-determined region. Then the new to-be-determined region is processed in the similar manner. The process stops when the absolute value of the difference between the two thresholds calculated between two iterations is less than a preset constant,and the last threshold calculated is the final threshold what we want. The paper not only give the segmentation result intuitively but also give the quantitative result of segmentation using the uniformity measure which is an image segmentation evaluation criteria. The experiment results show that the iterative method not only can get desired segmentation result intuitively but also that the uniformity measure calculated at each iteration is a monotone increasing sequence. And the experiment shows that the proposed method is not sensitive to parameter a.