Abstract:PSNR-oriented models are a critical class of super-resolution models with applications across various fields. However, these models tend to generate over-smoothed images, a problem that has been analyzed previously from the perspectives of models or loss functions, but without taking into account the impact of data properties. In this paper, we present a novel phenomenon that we term the center-oriented optimization (COO) problem, where a model's output converges towards the center point of similar high-resolution images, rather than towards the ground truth. We demonstrate that the strength of this problem is related to the uncertainty of data, which we quantify using entropy. We prove that as the entropy of high-resolution images increases, their center point will move further away from the clean image distribution, and the model will generate over-smoothed images. Implicitly optimizing the COO problem, perceptual-driven approaches such as perceptual loss, model structure optimization, or GAN-based methods can be viewed. We propose an explicit solution to the COO problem, called Detail Enhanced Contrastive Loss (DECLoss). DECLoss utilizes the clustering property of contrastive learning to directly reduce the variance of the potential high-resolution distribution and thereby decrease the entropy. We evaluate DECLoss on multiple super-resolution benchmarks and demonstrate that it improves the perceptual quality of PSNR-oriented models. Moreover, when applied to GAN-based methods, such as RaGAN, DECLoss helps to achieve state-of-the-art performance, such as 0.093 LPIPS with 24.51 PSNR on 4x downsampled Urban100, validating the effectiveness and generalization of our approach.

Soft Edge Smoothness Prior For Alpha Channel Super Resolution

Image Super Resolution Via Visual Prior Based Digital Image Characteristic

Improved Restoration Algorithm for Weakly Blurred and Strongly Noisy Image

A Single-Image Super-Resolution Method Based on Progressive-Iterative Approximation

Edge-preserving Smoothing for Image Decomposition Via a Hybrid Approach

EDGE-PRESERVING IMAGE SMOOTH ALGORITHM BASED ON LOCAL EXTREMA INTERPOLATION

Smoothing Algorithm with Edge-preserving by Extrema Constraints

Blind Image Deblurring via Superpixel Segmentation Prior

Edge-Aware Volume Smoothing Using L0 Gradient Minimization

Edge-Respecting Image Smoothing via Extrema Interpolation.

Structure-preserving image smoothing with semantic cues

Coarse Semantic Guided Alpha Matting Via Simultaneous Foreground and Background Estimation

Image smoothing based on global sparsity decomposition and a variable parameter

Neighborhood Regression for Edge-Preserving Image Super-Resolution.

Depth Super-Resolution Using Joint Adaptive Weighted Least Squares and Patching Gradient.

Enhancement of Contour Smoothness by Substitution of Interpolated Sub-Pixel Points for Edge Pixels

Image smoothing method based on global gradient sparsity and local relative gradient constraint optimization

Real-Time Image Smoothing Based on True Edges

Chromatic Aberration Correction Using Cross-Channel Prior in Shearlet Domain.

Edge Detection Method for High-Resolution Remote Sensing Imagery by Combining Superpixels with Dual-Threshold Edge Tracking

Uncovering the Over-smoothing Challenge in Image Super-Resolution: Entropy-based Quantification and Contrastive Optimization