一种改进的NAS-RIF图像盲复原算法

An Improved NAS-RIF Algorithm for Blind Image Restoration

针对模糊含噪退化图像的盲复原处理,文章以NAS-RIF算法为基础,对其存在的抑制噪声不理想的缺陷进行了相应改进.首先采用基于最小二乘支持向量机的去噪方法对退化图像进行预处理,在抑制噪声的同时保持图像的细节特征,进而在每次复原迭代过程中加入低通滤波环节,进一步减少噪声对代价函数收敛的影响,提高复原图像的信噪比.另外,对退化图像使用了阈值分割技术确定图像支撑域,保证复原的准确性.实验结果表明,改进的NAS-RIF算法抗噪声干扰的能力比原算法有显著的提高.

In order to achieve the blind restoration of the degraded images contaminated by noise,the erticle presents an improved NAS-RIF algorithm to overcome the drawbacks of the original algorithm. Firstly, an image denoising method based on least squares support vector ma- chine is introduced to filter the blurred image to increase the signal-to-noise ratio （SNR） which can effectively preserve the detail feature of the image. Then a low-pass filter in each iteration is used to alleviate the influence by noise on the cost function＇s convergence. In addition,an image segmentation technique which uses threshold to confirm the support of the image is introduced. The experiment shows that the improved algorithm has better restoration effect than the original one in the case of noise.