Super-Resolution with Deep Adaptive Image Resampling
This work addresses image super-resolution for computer vision applications, but it is incremental as it adapts traditional methods with deep learning rather than introducing a new paradigm.
The paper tackles single image super-resolution by revisiting interpolation-based methods with deep learning, using a CNN to estimate spatially variant interpolation kernels adaptively, achieving results on par with state-of-the-art methods and extending it to joint image filtering with similar performance.
Deep learning based methods have recently pushed the state-of-the-art on the problem of Single Image Super-Resolution (SISR). In this work, we revisit the more traditional interpolation-based methods, that were popular before, now with the help of deep learning. In particular, we propose to use a Convolutional Neural Network (CNN) to estimate spatially variant interpolation kernels and apply the estimated kernels adaptively to each position in the image. The whole model is trained in an end-to-end manner. We explore two ways to improve the results for the case of large upscaling factors, and propose a recursive extension of our basic model. This achieves results that are on par with state-of-the-art methods. We visualize the estimated adaptive interpolation kernels to gain more insight on the effectiveness of the proposed method. We also extend the method to the task of joint image filtering and again achieve state-of-the-art performance.