Zoom, Enhance! Measuring Surveillance GAN Up-sampling

TL;DR

This paper evaluates and compares CNN and GAN-based up-sampling techniques for security and surveillance images and videos, establishing DISTS as a superior quality metric for GAN up-sampling in this domain.

Contribution

It extends CNN and GAN applications to surveillance up-sampling and provides empirical evidence supporting DISTS as a better IQA metric.

Findings

GAN-based up-sampling outperforms traditional methods in surveillance images.

DISTS is identified as a more reliable IQA metric for GAN up-sampling.

Comparative analysis highlights strengths and weaknesses of CNN and GAN techniques.

Abstract

Deep Neural Networks have been very successfully used for many computer vision and pattern recognition applications. While Convolutional Neural Networks(CNNs) have shown the path to state of art image classifications, Generative Adversarial Networks or GANs have provided state of art capabilities in image generation. In this paper we extend the applications of CNNs and GANs to experiment with up-sampling techniques in the domains of security and surveillance. Through this work we evaluate, compare and contrast the state of art techniques in both CNN and GAN based image and video up-sampling in the surveillance domain. As a result of this study we also provide experimental evidence to establish DISTS as a stronger Image Quality Assessment(IQA) metric for comparing GAN Based Image Up-sampling in the surveillance domain.