摘要
图像去模糊本质上是求解一个病态问题。由于理论上图像均存在稀疏域的特点,l1凸松弛技术经常用来求解图像去模糊的病态问题。然而,在获取图像的实际过程中,不同类型的噪声可能会引入到模糊图像中。对于不同噪声污染的模糊图像,如果仍然采用同一模型进行图像去模糊,很难产生令人满意的结果。基于此,本文在分析噪声对模糊图像污染特点的基础上,提出采用不同的l1凸松弛模型去除图像模糊和噪声的方法。在所提的方法中,根据模糊图像的像素是全部还是部分被噪声污染,在l1凸松弛的优化模型中选用不同的保真项。实验结果验证了本文提出的基于噪声特点和l1凸松弛技术的图像去模糊方法的正确性和有效性。
Image deblurring is inherently to solve an ill-posed problem.The technique of l1convex relaxation is usually used to solve this ill-posed problem because any one image exists a sparse domain in theory.However,different types of noise may be introduced into the blur image in the practical process of acquiring the image.It is difficult to obtain satisfactory results if the same model is used to deblur the image for different degrees.In view of this,we firstly analyze the characteristics which the blur image is corrupted by noise,and then propose that different models of l1 convex relaxation are used to recover the image.In the proposed method,different fidelity terms are used in the optimization models of l1 convex relaxation according to all pixels or parts of the blur image corrupted by the noise.The experimental results verify the correctness and efficiency of the proposed method.
出处
《中国体视学与图像分析》
2011年第2期118-123,共6页
Chinese Journal of Stereology and Image Analysis
基金
国家自然科学基金(61033004
60736043
61070138)
国家教育部博士点基金资助项目(200807010004)
关键词
图像去模糊
l1凸松弛技术
噪声
稀疏表示
image deblurring
l1convex relaxation
noise
sparse representation
