摘要
针对传统自组织映射算法难以实现三维信号的非线性映射问题,提出三维自组织映射算法。采用二维输入层和三维输出层的神经网络结构,将邻域算法、竞争算法、学习算法和初始化算法的三维自组织映射算法应用于三维立体图像压缩编码。仿真实验表明,三维自组织映射算法的重构图像具有较好的峰值信噪比和主观品质。
The traditional self-organizing mapping algorithm is difficult to realize nonlinear mapping problem of three dimen-sional signal,the three-dimensional self-organizing mapping algorithm is put forward.Using two-dimensional input layer and three-dimensional output layer of neural network structure,the three-dimensional self-organizing mapping algorithm is researched including neighborhood algorithm,competition algorithm,learning algorithm and initialization algorithm,and it is applied to the three-dimensional image compression coding.The simulation experiments show that the reconstructed ima-ges of three-dimensional SOM algorithm have better peak signal-to-noise ratio and subj ective quality.
出处
《桂林电子科技大学学报》
2014年第2期120-125,共6页
Journal of Guilin University of Electronic Technology
基金
国家自然科学基金(61261035)
关键词
立体图像
邻域算法
竞争算法
学习算法
初始化算法
三维自组织映射
stereo image
neighborhood algorithm
competitive algorithm
learning algorithm
initialization algorithm
three-dimensional self-organizing mapping
