论文部分内容阅读
本文用改进的Kohonen神经网进行码书设计,并对误差图象再编码传送,形成多级矢量量化编码结构,算法简便,并可在码率低于1.0bpp的情况下得到较好的图象质量。
In this paper, an improved Kohonen neural network is used to design the codebook, and the error image is re-encoded and transmitted to form a multi-level vector quantization coding structure. The algorithm is simple and can obtain a better image with the code rate lower than 1.0bpp quality.