自适应分形矢量量化编码

A Self-Adaptive Algorithm of Fractal Vector Quantization for Image Coding

提出了一种自适应分形矢量量化编码方法。对图像进行自适应四叉树分割 ,并构造粗糙的均值图像。误差图像进行分形矢量量化编码时 ,使用设计的自适应二维维纳数字滤波器 ,对收缩的均值图像进行滤波后 ,可构造好的码书。实验证明 ,本方法码书不需外部训练 ,解码不需要迭代 ,可以改善重建图像的视觉质量 ,使压缩比和PSNR都有明显提高。

A self-adaptive algorithm of fractal vector quantization for image coding is proposed. A self-adaptive quad-tree is used to partition the original image, then structured coarsely mean value image. When coding the erroneous image by fractal vector quantization, a self-adaptive 2-D wiener digital filter to filtrate the contracted mean value image, which can result into a good quality codebook. Computer simulation of encoding and decoding show that no training is needed to obtain the codebook in encoding, and decoding can be done without iteration. The method makes better the visual quality of the reconstructed image and improves the compression ration and PSNR.