Focus on Objects: An Image Super-Resolution Model for Region of Interest

Image super-resolution (SR) can reconstruct a low-resolution (LR) image into a high-resolution (HR) image to provide an excellent visual experience. However, this task is pixel-wise, and for images with an ordinary resolution, its processing speed is inefficient, which is challenging to meet the needs of practical applications. People only focus on local regions in the image and step back to super-resolve these regions of interest, which can significantly reduce the computational complexity. We propose an image SR model that only focuses on the region of interest in one image, effectively uses people’s visual attention and significantly reduces computing power consumption.