The Use of Double Moment Based Descriptors to Speed up FIC

Abstract

In this research, an improved Partitioned Iterated Function System (PIFS) scheme is introduced to speed up the encoding stage time in Fractal Image Compression (FIC). The proposed system is based on using both symmetry prediction and blocks indexing to speed up the block matching process; these two methods have been used to manage the way of selecting the suitable domain to represent the range block. For block indexing the moment features have been used for determining double block descriptors, each one is affine invariant. The two block descriptors are utilized in a combined indexing scheme to classify the blocks of both range and domain pool. So, the block indexing method is introduced to filter the domain blocks, and keeps only those domain blocks have similar block indices with that of the mapped range block which will be approximately represented using affine mapping. The symmetry predictor is used to reduce the number of isometric trails from 8 to one trail .The research also includes the implementation of the use of single moment descriptor to speed up FIC for comparison. The test results indicated that the proposed improvement has reduced the required encoding time to 0.25 second, and the attain compression ratio is 7.99 without making significant degradation in image quality.