An improved algorithm with importance weight value based on super-resolution through neighbor embedding

Based on the research on super-resolution algorithm of neighbor embedding (SRNE), this paper proposes an improved algorithm, namely, super-resolution neighbor embedding with importance weight value (NEIWV) to improve the image resolution. As in dealing with special types of images, especially medical images, the images' color and texture features in the target region play important roles. However, the importance levels of these characteristics are different; we calculate the importance weight value and then embed the value into the SRNE. Each feature of all image patches is represented with a feature vector. Through selecting the feature vector, we restore the information from the high and low frequencies information of the image as much as possible. In this algorithm we choose the Gaussian intensity vector and first-order gradient vector. The experiment results show that the improved algorithm has a good effect in raising the image resolution and eliminating noise.