一种基于DCT边缘模式分类的神经网络图象编码方法
An Image Coding Method Based on DCT Edge Classification Using Neural Networks
-
摘要: 对神经网络应用于图象数据的压缩编码进行了研究和分析,提出了一种基于DCT边缘模式分类的神经网络图象编码方法,分析了网络中隐蔽层节点数与网络的记忆能力之间的关系,在网络的学习训练过程中,采用了一种多层前向网络的快速学习算法,并对其计算进行了进一步简化.初步实验结果表明,采用该方法较好地保持了图象中的边缘信息.Abstract: This paper dicussses the application of Neural Networks on image data compression.According to its efficient information storage and data compression ability, an image data compression system using the Back-Propagation networks and a fast training algorithm of multi-layer neural networks are proposed. A model of image data compression based on DCT edge pattern classification is introduced. The experiment results demonstrats a satisfying performance on edge feature preserving in the image.