基于图的数字全变差模型及其带噪图像任意精度放大
Digital Total Variation Model for Image Zoom-In with Arbitrary Accuracy
-
摘要: 分析了利用Sobolev空间Tikhonov正则化模型对带噪图像进行放大的不足,基于图像的修复模型,提出带噪图像放大的数字全变差模型.利用有向图构造出兼顾噪声去除和图像放大的数字TV滤波器,并利用该滤波器提出一种新颖的图像放大算法.作为算法对比,利用Sobolev空间Tikhonov正则化模型,提出相对应的数字Tikhonov 放大算法.结果表明:数字TV放大算法明显优于数字Tikhonov放大算法,不仅较好地抑制了噪声的影响,而且使得任意精度放大的图像边缘清晰、过渡自然,特别适合于目标边缘明显的一类非纹理的医学图像的放大.Abstract: Shortcomings of Tikhonov regularization model in Sobolev space for image zoom in are analyzed and a digital total variation model with image inpainting for noisy image zoom-in is presented. Described by a directed graph, a digital TV algorithm for noisy image zoom-in is implemented. For comparison, the digital Tikhonov zoom-in algorithm is also introduced. Tests show that the digital TV algorithm achieves better performance, it can zoom an image with arbitrary accuracy and keep the edges as clear and natural as possible. The algorithm is very suitable for non-texture medical image.
下载: