摘要
本文利用牛顿前向插值多项式构建一种联想记忆型神经网络,并基于联想记忆提出一种彩色图像的自适应预测无损编码方法。针对彩色图像帧间相关性较大的特点,首先用R,G,B-G替代R,G,B以减小帧间相关性,然后对R 单独进行预测编码,对C,B—G进行二维矢量预测编码进一步减小空间相关和帧间相关,最后对预测残差进行算术编码。实验证明该算法是可行性的和有效性的,对测试图像的压缩效果优于JPEG-LS标准。
A novel Associative Memory System（NFI-AMS） based on Newton＇ s Forward Interpolation is presented in this paper. Combining NFI -AMS and arithmetic coding, an algorithm of adaptive prediction for image lossless compression is proposed. For a given color image, first components R, G, B-G are used to replace R, G, B to reduce spectral correlation, then component R is coded independently and G, B-G are coded jointly to reduce spectral and spatial correlation simultaneously, at last arithmetic coding is exploited to code with errors between predictive value and source image. Experiments give better results than JPEG-LS and show that the method is effective for image compression.
出处
《信号处理》
CSCD
北大核心
2006年第2期136-138,共3页
Journal of Signal Processing
关键词
联想记忆
神经网络
无损压缩
自适应预测编码
Associative Memory System
Neural Network
Lossless Compression
Adaptive prediction coding
