基于BWT和FVQ的极低比特率图像编码算法

BWT AND FVQ BASED VERY LOW BIT RATE IMAGE CODING ALGORITHM

该文提出了一种基于双正交小波变换(BWT)和模糊矢量量化(FVQ)的极低比特率图像编码算法.该算法通过构造符合图像小波变换系数特征的跨频带矢量,充分利用了不同频带小波系数之间的相关性,有效地提高了图像的编码效率和重构质量.该算法采用非线性插补矢量量化(NLIVQ)的思想,从大维数矢量中提取小维数的特征矢量,并提出了一种新的模糊矢量量化方法——渐进构造模糊聚类(PCFC)算法用于特征矢量的量化,从而大大提高了矢量量化的速度和码书质量.实验结果证明,该算法在比特率为0.172bpp的条件下仍能获得PSNR>30dB的高质量重构图像。

A biorthogonal wavelet transform (BWT) and fuzzy vector quantization (FVQ) based very low bit rate image coding algorithm is proposed. The correlation of the wavelet coefficients in different frequency bands is fully exploited through constructing the band-cross vector, so the high coding efficiency and reconstructed image quality are obtained simultaneously. In addition, a hybrid vector quantization (VQ) scheme is presented to improve the performance of VQ, which combines the non-linear interpolated vector quantization (NLIVQ) technique with a novel progressive constructive fuzzy clustering algorithm. Simulation results demonstrate that the reconstruction quality is higher than 30dB at a very low bit rate of 0.172bpp.