Fast classification method for fractal image compression

In this paper, a fast algorithm is developed which reduces the searching space for Fractal image coding. The basic idea is to classify the domain pool into three classes, non-edged class, horizontal/vertical class and the diagonal class. For each given range block the property is computed first to determine which class it belongs. Then one only has to search from the corresponding class in the domain pool to find the best match. The classification operation is performed only according to the lowest frequency coefficients of the given block in the horizontal and vertical directions, in which the frequency data is computed from Discrete Cosine Transform (DCT). There main advantages for this classification scheme are that the classification mechanism is simple and the DCT algorithm is easy to implement. A simulation shows that, the proposed fast algorithm is about 2 times faster than the baseline method while the quality of the retrieved image is almost the same.