Fast Fractal Image Compression based on Fisher´s classification scheme

Image compression and retrieval techniques are essential for visual database management in an efficient manner. These techniques may vary for each application. Fractal Image Compression is one of the best techniques for natural and still images. In this method an image is divided into non overlapping range blocks and overlapping domain blocks. The domain blocks are larger than the range blocks in size and number. All the domain blocks are collectively called as domain pool. Size of the domain pool determines the complexity of encoding phase. Each range block is encoded based on affine similarity between the domain blocks. The best matched domain block for each range block is given by Absolute Value of Pearson´s Correlation Coefficient. Regardless of various advantages offered by fractal compression, such as high speed, high bit rate, high decompression and resolution independence, the major disadvantage is the high computational cost of the coding phase. This paper proposes two methods to reduce the complexity of the image coding phase. The first method classifies the domain pool into three classes with Fisher´s classification technique and in the second method, a specific number of blocks of one class are considered for Fast Fractal Image Compression.