Impact of a Sharpness Based Loss Function for Removing Out-of-Focus Blur

Abstract

Recent research has explored complex loss functions for deblurring. In this work, we explore the impact of a previously introduced loss function - Q which explicitly addresses sharpness and employ it to fine-tune State-of-the-Art (SOTA) deblurring models. Standard image quality metrics such as PSNR or SSIM do not distinguish sharpness from ringing. Therefore, we propose a novel full-reference image quality metric Omega that combines PSNR with Q. This metric is sensitive to ringing artefacts, but not to a slight increase in sharpness, thus making it a fair metric for comparing restorations from deblurring mechanisms. Our approach shows an increase of 15 percent in sharpness (Q) and up to 10 percent in Omega over the use of standard losses.

Categories