A Cosine Network for Image Super-Resolution
Chunwei Tian, Chengyuan Zhang, Bob Zhang, Zhiwu Li, C. L. Philip Chen, David Zhang
TL;DR
We address single-image super-resolution by combining a heterogeneous network design with a cosine-based training strategy. The CSRNet architecture uses odd/even enhancement blocks to capture complementary structural information while preserving gradients through residual paths, and it employs a cosine annealing schedule to avoid local minima during training. Empirical results on standard SR benchmarks show competitive PSNR/SSIM against state-of-the-art methods, with ablations validating the effectiveness of the OEB/EEB modules and the cosine optimizer. This approach offers a robust, scalable SR solution with potential for adaptive and mobile deployment.
Abstract
Deep convolutional neural networks can use hierarchical information to progressively extract structural information to recover high-quality images. However, preserving the effectiveness of the obtained structural information is important in image super-resolution. In this paper, we propose a cosine network for image super-resolution (CSRNet) by improving a network architecture and optimizing the training strategy. To extract complementary homologous structural information, odd and even heterogeneous blocks are designed to enlarge the architectural differences and improve the performance of image super-resolution. Combining linear and non-linear structural information can overcome the drawback of homologous information and enhance the robustness of the obtained structural information in image super-resolution. Taking into account the local minimum of gradient descent, a cosine annealing mechanism is used to optimize the training procedure by performing warm restarts and adjusting the learning rate. Experimental results illustrate that the proposed CSRNet is competitive with state-of-the-art methods in image super-resolution.