CFA图像无损压缩神经网络预测器的设计  

Design of Neural Network predictor for lossless compression of CFA image

作　　者：张文爱[1] 赵建刚[1] 程永强[1] 

机构地区：[1]太原理工大学信息工程学院,太原030024

出　　处：《计算机工程与应用》2009年第5期174-176,共3页Computer Engineering and Applications

摘　　要：论文将神经网络引入预测器设计中,实现了一个用于贝尔模板自然图像无损压缩的神经网络预测器。该预测器含有两个隐含层,具有很强的非线性预测能力,可直接对贝尔模板类型的图像进行无损压缩。为了提高预测精度,在CFA数据的因果邻域像素中实现了边缘检测,并将其应用到预测器中,实验表明该边缘检测方法简单有效。In this paper,neural network is introduced into the design of predictor,and a neural network predictor for lossless compression of Bayer pattern image is implemented.The predictor which consists of two hidden layers is very suitable for nonlinear prediction,and can be used directly in the lossless compression of CFA raw data.In order to increase the precision of the prediction,an edge detection algorithm is proposed.Experiments show that the algorithm is simple and efficient.

关 键 词：神经网络 预测编码 贝尔模板 边缘检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]