Fractal Image Compression Based on Wavelet Transform
ZB Zhang,GX Zhu,YT Zhu
DOI: https://doi.org/10.1117/12.271719
1997-01-01
Abstract:It is well known that images can be greatly compressed by exploiting the self-similar redundancies. In this paper, the self-similarities of wavelet transform are analyzed, and it is discovered that corresponding subbands of different scale detail signals are similar. An image coding method is proposed according to this property. The typical self-affirm transform is modified such that it is adapted to DWT coefficient encoding. An adaptive quantization method of the transform parameters s, is given. Firstly, a J-order discrete wavelet transform of the original image, denoted by LL0, is performed. That is, LL is decomposed into LLj + 1, LHj + 1, HL$_j + 1, for 0 <EQ j <EQ J - 1. Secondly, LLJ$. is encoded based on DCT. Thirdly, HL(subscript J LHJ and HHJ are quantized and run- length coded. Fourthly, HLj, LHj and HHj for 1 <EQ j <EQ J - 1, are encoded with modified self- similar transforms. HLj, LHj, and HHj are divided into non-overlapping range blocks. For each range block Ri (epsilon HLj, a domain block Dj (epsilon) HLj + 1, which best matches Rj, is found, and the parameter s1 of the corresponding transform is determined and adaptively quantized. Several kinds of images are compressed with this method. Experimental results demonstrate that this method can compress images significantly while keeping a very good fidelity. Besides, the algorithm is faster than typical fractal image coding methods because less searching is needed.