Research On Fractal Image Compression Based On Numerical Classification

With the internet growing at an exponential rate, the amount of people transferring digital data from on site to another has increased dramatically. There is now, more than ever, a need for quick data transfer methods and more efficient use of memory space. Unfortunately, images are known for requiring significantly large amount of memory and hence are not transmittable quickly. This is where fractal image compression techniques come to the rescue. Such techniques allow one to store an image with much less memory than it would normally require, hence allowing it to be transmitted more quickly.The existing fractal image compression algorithm based on the grayscale average classification of divide the image into three categories, and each of the categories has24small classes, that is to say, the image can be divided into72classes totally. This algorithm has the problem that long encoding time, poor matching accuracy and long search time and so on.This thesis mainly discusses the fractal image compression algorithm based on the numerical classification and introduces the basic theory of fractal image compression, several basic models of classification, the algorithm classification of the fractal image compression, and the existing fractal image compression algorithm based on the grayscale average classification. In view of the defects of the existing fractal image compression algorithm based on the grayscale average classification, a fractal image compression algorithm based on the numerical classification is proposed in this thesis. The algorithm includes the image segmentation method, image rotate-affine transformation operation as well as the calculation of error. In this thesis, the main work includes the following two aspects:1.In the process of image classification, this thesis adopted the "sum-sampling" transformation rather than "average-sampling" transformation, the reason is "average-sampling" transformation has one more division operation relative to the "sum-sampling" transformation, which will cause larger amount of calculation and lower accuracy.2.The improved method in this thesis is on the basis of the Fisher classification method, sorting the grey value, then divide the gray value range into eight grade. According to the different corresponding relations of the sorted grey value among the eight grade, we can divide the image into a considerable amount of classes, so that to improve the search speed, and solve the problem of low coding speed. The simulation experimental results show that the improved method proposed in this thesis accelerate the encoding speed effectively, reduce the encoding time on the premise of guarantee the quality of image.The simulation experimental results show that the improved method mentioned in the article speed up the encoding speed effectively, reduce the encoding time on the premise of guarantee the quality of image.