基于聚类和回归的超分辨率方法

基于实例的超分辨算法在超分辨重建技术中得到了广泛的应用,如邻域嵌入算法。这个算法的理论基础建立在流形假设上,而流形假设往往并不是成立。为了避免流形假设,提出了基于聚类和回归的超分辨率方法。由于图像的信息较为复杂,因此首先将图像样本块进行聚类。然后,在高分辨图像块与低分辨图像块中学习局部回归函数。最后应用回归函数进行超分辨重建。将仿真实验结果与其他算法进行比较,证明了该算法的可行性和有效性。

Super-resolution via Clustering and Regression

Example-based super-resolution algorithm has been widely used in the super-resolution reconstruction technology,such as neighbor embedding algorithm.These algorithms are based on manifold assumption,while manifold assumption does not always hold.In order to avoid the assumption,a super-resolution method based on cluster and regression is presented.Because of the complexity of image information,firstly,the image patches are clustered.Then,a local function is learned between high-resolution image patches and ...