Image Interpolation using Subband Pseudo 2-D HMM

T. Nagai, A. Kurematsu (Japan), and T. Nguyen (USA)


Image Interpolation, Object-Specific Knowledge, Pseudo 2D HMM


In this paper, a novel approach to the image interpola tion is presented. Our proposed approach is based on a priori knowledge about a category, to which the input image belongs. The knowledge is acquired by learning from a number of training samples for every category in the training phase. The interpolation is carried out by recognizing the category and using the knowledge of the category. To achieve this, we use the Hidden Markov Model (HMM) framework, which models the correspondence between an input image and its high resolution version. We also apply our algorithm to im age super-resolution, which combines multiple input images. The experimental results show the effective ness of the proposed method.

Important Links:

Go Back