Input Image-Based Dictionary Formation in Super-Resolution for Online Image Streaming

Abstract

Real-time images are very much desirable in different real-life applications such as defense, medical, satellite, tv, Internet. Construction of relevant dictionary for the training of computer during the prediction of high-resolution image from low-resolution images in real-time requires high mathematical computations. This paper presents a new way of dictionary formation for learning-based super-resolution of real-time streaming of images. Here, the dictionary is formed with help of images that are similar to the input image in terms of structural similarity score. It helps in reducing memory requirement for the dictionary formation. Further, to speed up the process, a technique based on similarity score is proposed for updating of the dictionary. This involves the comparison of current image in the input sequence with the present reference image that was used for dictionary formation. Efficacy of the algorithm is shown through extensive simulations.