低失真敏感测量度图像编码方法研究  

Research on Method of Image Coding With Low Distortion and Sensibility Gauge Property

作　　者：张翼飞[1] 高彦卿[2] 

机构地区：[1]河南城建学院计算机科学与工程学院,河南平顶山467036 [2]南阳理工学院软件学院,河南南阳473004

出　　处：《科技通报》2014年第2期194-196,共3页Bulletin of Science and Technology

摘　　要：传统的自组织神经网络系统进行图像量化处理和编码中,算法对于码书初始值选择具有较强的敏感性,从而导致该量化编码算法对训练序列具有较大的依赖性。为此提出基于自组织神经映射网络系统的低失真敏感测量度图像编码图像量化编码方案,主要是针对初始码书的较强敏感性而改进的,以便改进码书的性能和训练效率。改进的方法中通过设置一个响应的频率计数器。由失真的敏感参数对失真测度进行调整,有效降低失真程度,从而减少了已被用作响应码氏的码氏再度成为响应码氏的概率,实现了对码书性能和数量之间的矛盾进行有效地平衡折中。训练效果和仿真结果表明,方法改进效果明显,其峰值信噪比相比提高了3.75 dB,图像还原效果真实有效,能有效适用于对图像向量量化和编码。In the traditional self -organizing feature mapping algorithm network system algorithm, the performance of the algorithm was much sensitive to the initial value of the codebook. An improved image coding method with low distortion and low sensitive gauge property was proposed based on the self-organizing feature mapping network. The distortion de-gree was decreased effectively. The probability of the coding vector which had been used as the response vector and was going to be taken as the response vector again was decreased finally, and the contradiction of the performance and the quantity of the coding book was balanced. Result shows that the performance of the improvement is obvious, and the peak signal-Noise ratio is improves with 3.75 dB, the performance of image restoration is true and effective, and it can be applied in the image vector quantization and image coding in practice.

关 键 词：自组织特征映射 敏感性 向量量化 图像编码 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]