Loss Functions for Neural Networks for Image Processing

TL;DR

This paper explores alternative loss functions for neural networks in image processing, emphasizing perceptually-motivated losses to improve image quality as judged by humans, and introduces a novel differentiable error function.

Contribution

It highlights the importance of loss function choice in image restoration and proposes a new differentiable loss that enhances results without changing network architecture.

Findings

Perceptually-motivated losses outperform L2 in image restoration tasks.

The proposed differentiable error function improves image quality significantly.

Better loss functions lead to higher human-perceived image quality.

Abstract

Neural networks are becoming central in several areas of computer vision and image processing and different architectures have been proposed to solve specific problems. The impact of the loss layer of neural networks, however, has not received much attention in the context of image processing: the default and virtually only choice is L2. In this paper, we bring attention to alternative choices for image restoration. In particular, we show the importance of perceptually-motivated losses when the resulting image is to be evaluated by a human observer. We compare the performance of several losses, and propose a novel, differentiable error function. We show that the quality of the results improves significantly with better loss functions, even when the network architecture is left unchanged.