On the Importance of Denoising when Learning to Compress Images
This addresses the issue of inefficient compression for noisy images in photography, offering a more efficient solution, though it is incremental as it builds on existing compression and denoising methods.
The paper tackles the problem of image compression in the presence of noise by learning denoising within the codec training, achieving better rate-distortion results with nearly one order of magnitude fewer GMac operations compared to compression-only or separate denoising-then-compression models.
Image noise is ubiquitous in photography. However, image noise is not compressible nor desirable, thus attempting to convey the noise in compressed image bitstreams yields sub-par results in both rate and distortion. We propose to explicitly learn the image denoising task when training a codec. Therefore, we leverage the Natural Image Noise Dataset, which offers a wide variety of scenes captured with various ISO numbers, leading to different noise levels, including insignificant ones. Given this training set, we supervise the codec with noisy-clean image pairs, and show that a single model trained based on a mixture of images with variable noise levels appears to yield best-in-class results with both noisy and clean images, achieving better rate-distortion than a compression-only model or even than a pair of denoising-then-compression models with almost one order of magnitude fewer GMac operations.