Towards Bidirectional Arbitrary Image Rescaling: Joint Optimization and Cycle Idempotence
This work addresses the need for more flexible and robust image rescaling models in real-world applications, representing a novel approach but with incremental advancements in the field.
The paper tackles the problem of arbitrary image rescaling (both upscaling and downscaling) by proposing a unified model that jointly optimizes both directions, achieving significant performance improvements over existing arbitrary upscaling models and maintaining visual quality in downscaled images, with robustness in cycle idempotence tests.
Deep learning based single image super-resolution models have been widely studied and superb results are achieved in upscaling low-resolution images with fixed scale factor and downscaling degradation kernel. To improve real world applicability of such models, there are growing interests to develop models optimized for arbitrary upscaling factors. Our proposed method is the first to treat arbitrary rescaling, both upscaling and downscaling, as one unified process. Using joint optimization of both directions, the proposed model is able to learn upscaling and downscaling simultaneously and achieve bidirectional arbitrary image rescaling. It improves the performance of current arbitrary upscaling models by a large margin while at the same time learns to maintain visual perception quality in downscaled images. The proposed model is further shown to be robust in cycle idempotence test, free of severe degradations in reconstruction accuracy when the downscaling-to-upscaling cycle is applied repetitively. This robustness is beneficial for image rescaling in the wild when this cycle could be applied to one image for multiple times. It also performs well on tests with arbitrary large scales and asymmetric scales, even when the model is not trained with such tasks. Extensive experiments are conducted to demonstrate the superior performance of our model.