Learning-based super-resolution image reconstruction on multi-core processor

Super-resolution image reconstruction is an important technology in many image processing areas such as image sensing, medical imaging, satellite imaging, and television signal conversion. It is also being used as a unique selling point for a recent consumer HDTV set equipped with a multi-core processor. Among the various super-resolution methods, the learning-based method is one of the most promising solutions. However, this method is difficult to implement in real time because of the computational time required for searching the large database of reference images. In this paper, we propose a new learning-based superresolution method that utilizes total variation (TV) regularization. We obtain excellent image quality improvement and a large reduction in the computational time with this method. Our method was implemented on a multi-core processor to examine the possibility of real-time processing. The method enables the adoption of learning-based super-resolution for current HDTV sets equipped with multi-core processors as well as for the next generation HDTVs with 4 K × 2 K panels.