摘要
针对当前图像降噪算法难以有效区分纹理区域边缘和细节,且其噪声大小估计不准确,使其降噪质量不佳等不足,提出一种基于噪声大小估计耦合PCA的图像降噪算法。根据图像块的梯度矩阵的纹理强度和统计信息,从图像中选择无高频成分的低等级块,利用PCA技术计算图像块的协方差矩阵的特征值,利用最小特征值表征图像初始噪声;引入噪声估计函数,通过不断迭代估计函数,直到计算的真实噪声不变为止;根据图像真实噪声和场景复杂性,调整噪声大小,利用调整后的噪声大小作为降噪标准。实验结果表明,与当前降噪算法相比,所提算法在边缘和丰富纹理区域具有更好的降噪效果和更高的稳定性。
For the noise images, it is difficult to distinguish the edges and details, and the noise size estimation is not accurate, which se- riously affect the performance of image denoising algorithm. The image denoising algorithm based on noise size estimation coupled with PCA was proposed. According to the texture and statistical information of the gradient matrix of the image block, the low grade block which did not contain the high frequency component was chosen from the image, and the PCA technique was used to calculate the eigen- values of the covariance matrix of the image block, and the minimum eigenvalue of the covariance matrix was used to characterize the ini- tial noise of the image. The noise estimation function was introduced, by constantly iterating estimation function, the real noise was cal- culateck According to the image real noise and the complexity of the scene, the size of the noise was adjusted, and the adjusted noise was used as the standard for denoising. Experimental results show that comparing with the current commonly denoising algorithms, this algo- rithm has good denoising effects, and it is more stable in the edge and rich texture region.
出处
《计算机工程与设计》
北大核心
2017年第4期959-964,1109,共7页
Computer Engineering and Design
基金
国家青年基金项目(61100132)
关键词
图像降噪
噪声估计
主成分分析
梯度矩阵
降噪标准
image denoising
noise estimation
PCA
gradient matrix
noise reduction standard
