Grey is the new RGB: How good is GAN-based image colorization for image compression?
- PDF / 2,549,809 Bytes
- 17 Pages / 439.642 x 666.49 pts Page_size
- 36 Downloads / 216 Views
Grey is the new RGB: How good is GAN-based image colorization for image compression? Aroosh Fatima1,2 · Wajahat Hussain2,3 · Shahzad Rasool1,2 Received: 16 December 2019 / Revised: 15 July 2020 / Accepted: 9 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract GAN-based image colorization techniques are capable of producing highly realistic color in real-time. Subjective assessment of these approaches has demonstrated that humans are unable to differentiate between a true RGB image and a colorized image. In this work, we evaluate the fidelity of such colorization and for the first time analyze the GAN-based image colorization scheme in the context of image compression. Our analysis shows that the palette (set of colors) recommended by the GAN-based framework is very limited even for highly realistic interactive colorization. We propose two novel methods of automatic palette generation that allows for the GAN-based framework to be useful for image compression. We demonstrate that provided true colors at a few pixel locations, GAN-based approach results in good spread of color to other image regions. Subjective analysis on a number of public datasets shows that the current system has low fidelity but performs better than JPEG at low data rate regimes. Keywords Image compression · Image colorization · Image quality assessment · Deep learning · GAN
1 Introduction The new normal dictated by the pandemic indicates that human civilization will migrate to digital space. The new office will be work from home. The new class will be online class. Shahzad Rasool
[email protected] Aroosh Fatima [email protected] Wajahat Hussain [email protected] 1
Research Centre for Modelling and Simulation (RCMS), Islamabad, Pakistan
2
National University of Sciences and Technology (NUST), Islamabad, Pakistan
3
School of Electrical Engineering and Computer Science (SEECS), Islamabad, Pakistan
Multimedia Tools and Applications
This migration has increased the humanity’s data requirements. How to meet these data requirements? This question refuels the interest in compression of images and videos for efficient storage and transmission. One way to achieve compression of images and videos is to transmit only the structural information and discard the color information to save storage space and reduce bandwidth requirements. However, it has been demonstrated that color information is important as more striking colors lead to increase in the popularity and memorability of an image [6, 8]. With the advent of deep learning and artificial intelligence a renewed interest has been seen towards image colorization. GAN-based colorization schemes that produce realistic images in real time with minimal human intervention have been proposed [14, 20]. Figure 1 shows an example of an image colorized, in real time, using deep learning [20]. This iconic image of Einstein was captured in greyscale. However, the automatic colorization (first row in Fig. 1), using deep lear
Data Loading...
