A Novel SVM-Based Blind Super-Resolution Algorithm
Jianping Qiao, Ju Liu, Caihua Zhao · The 2006 IEEE International Joint Conference on Neural Network Proceedings · 2006
In this paper, we propose a novel Support Vector Machines (SVM)-based method of blind Super-Resolution (SR) image restoration. First, a blur identification method is proposed to identify the blur parameter of the acquisition system from the compressed/uncompressed low-resolution image. In this method, SVM is used to classify feature vectors extracted from the training images by Sobel operator and local variance, the acquired mapping between the vectors and corresponding blur parameter provides the identification of the blur. After blur identification, a super-resolution image is reconstructed by a learning-based method in which Image Euclidean Distance (IMED) is used as a distance measurement during patch matching and different color channels are treated unequally to reduce the computation complexity. Experiments on both synthetic and real images demonstrate the effectiveness and robustness of our method.