Bayesian Optimal Latent Projection for Noisy Image Restoration

Ziqiang Shi, Rujie Liu, Jun Takahashi, Takuma Yamamoto; Proceedings of the Winter Conference on Applications of Computer Vision (WACV), 2025, pp. 2799-2807

In recent years image restoration using large-scale latent diffusion generative models (DGM) has attracted increasing attention and achieved significant progress. Most of these latent DGM-based image restoration methods require predicting the original clean image in each iteration which is then used to estimate the image for the next iteration. However these predicted original clean images are often inaccurate leading to errors in the subsequent image estimation. In other words there is a significant deviation between the final sampling restoration trajectory and the ground truth trajectory. The purpose of this paper is to narrow the gap between these two trajectories and enhance the performance of image restoration. We propose the Bayesian Optimal Latent Projection (BOLP) algorithm which identifies the optimal random noise within the Gaussian distribution to iteratively correct the estimated image at each step thereby minimizing the distance to the ground truth image. Experiments in deblurring super-resolution and inpainting on FFHQ and ImageNet datasets demonstrate that the BOLP outperforms the previously established best algorithms and sets a new state of the art.