灰度分布基于规范式描述的边缘模型基图像编码

Edge-based Model Video Coding Using Normal Forms Representation on the Gray Value Distribution

提出了一种称为边缘模型基的低比特率图像编码方法，其基本特点是：1）建立一种基于图像边缘信息的活动图像信源模型，按照图像中物体的结构将图像划分为激励区和缓变区（平滑区）。2）用三阶规范式拟合图像激励区的灰度分布，以“激程扫描─规范式拟合”表达图像的缓变区。3）引入了模型匹配与广义运动分析概念，并用干帧间图像预测编码。实验结果表明：在保持重建图像质量的前提下，每像素的比特数可压缩至0.075。

In this paper, we proposed a new scheme on low bit rate video coding which called Edge-based Model.There are three new ideas:i)A source model based on edge information is introduced which try to divide image infomation into sensitized region and smooth region by objects'structures.ii) The gray value distribution within sensitized region is fitted and represented by 3degree normal form polynomials while smooth region is described by a program of 'Run-lengthscanning and normal forms fitting'.iii) New concepts of model match and general motion analysis are introduced and used for prediction coding of inter-frames.Experimental results show that the subjective quality of reconstructed pictures are excellent at 0.075 bit-per-pei.