Memory augment is All You Need for image restoration
This addresses the problem of lack of transparency and aesthetics in CNN-based image restoration methods, though it appears incremental as it builds on existing contrastive learning and memory techniques.
The paper tackles image restoration by proposing MemoryNet, which uses a three-granularity memory layer and contrastive learning to improve performance on tasks like deraining, deshadowing, and deblurring, achieving significant PSNR and SSIM gains on multiple datasets.
Image restoration is a low-level vision task, most CNN methods are designed as a black box, lacking transparency and internal aesthetics. Although some methods combining traditional optimization algorithms with DNNs have been proposed, they all have some limitations. In this paper, we propose a three-granularity memory layer and contrast learning named MemoryNet, specifically, dividing the samples into positive, negative, and actual three samples for contrastive learning, where the memory layer is able to preserve the deep features of the image and the contrastive learning converges the learned features to balance. Experiments on Derain/Deshadow/Deblur task demonstrate that these methods are effective in improving restoration performance. In addition, this paper's model obtains significant PSNR, SSIM gain on three datasets with different degradation types, which is a strong proof that the recovered images are perceptually realistic. The source code of MemoryNet can be obtained from https://github.com/zhangbaijin/MemoryNet