A study of deep perceptual metrics for image quality assessment
This work addresses the problem of accurate image quality assessment for applications like image processing and computer vision, but it is incremental as it builds on existing deep perceptual metrics.
The study tackled the inefficiency of existing metrics for measuring similarity in highly distorted images by empirically investigating deep perceptual metrics for Image Quality Assessment, resulting in a multi-resolution perceptual metric (MR-Perceptual) that outperforms standard metrics on tasks with varying image deformations.
Several metrics exist to quantify the similarity between images, but they are inefficient when it comes to measure the similarity of highly distorted images. In this work, we propose to empirically investigate perceptual metrics based on deep neural networks for tackling the Image Quality Assessment (IQA) task. We study deep perceptual metrics according to different hyperparameters like the network's architecture or training procedure. Finally, we propose our multi-resolution perceptual metric (MR-Perceptual), that allows us to aggregate perceptual information at different resolutions and outperforms standard perceptual metrics on IQA tasks with varying image deformations. Our code is available at https://github.com/ENSTA-U2IS/MR_perceptual