Computer Graphics Laboratory ETH Zurich

ETH

Perceptually Based Downscaling of Images

A. C. Oztireli, M. Gross

Proceedings of ACM SIGGRAPH (Los Angeles, California, USA, August 9-13, 2015), ACM Transactions on Graphics, vol. 34, no. 4, pp. 77:1-77:10
[Abstract] [BibTeX] [PDF] [Video]

Abstract

We propose a perceptually based method for downscaling images that provides a better apparent depiction of the input image. We formulate image downscaling as an optimization problem where the difference between the input and output images is measured using a widely adopted perceptual image quality metric. The downscaled images retain perceptually important features and details, resulting in an accurate and spatio-temporally consistent representation of the high resolution input. We derive the solution of the optimization problem in closed-form, which leads to a simple, efficient and parallelizable implementation with sums and convolutions. The algorithm has running times similar to linear filtering and is orders of magnitude faster than the state-of-the-art for image downscaling. We validate the effectiveness of the technique with extensive tests on many images, video, and by performing a user study, which indicates a clear preference for the results of the new algorithm.

[Download Video]

@article{Oztireli15Perceptual,
author = {\"{O}ztireli, A. Cengiz and Gross, Markus},
title = {Perceptually Based Downscaling of Images},
journal = {ACM Trans. Graph.},
issue_date = {August 2015},
volume = {34},
number = {4},
month = jul,
year = {2015},
issn = {0730-0301},
pages = {77:1--77:10},
articleno = {77},
numpages = {10},
url = {http://doi.acm.org/10.1145/2766891},
doi = {10.1145/2766891},
acmid = {2766891},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {downscaling, images, perceptual, structural similarity, unsharp masking, video},
}
[Download BibTeX]

Overview

We propose a perceptually based method for downscaling images that provides a better apparent depiction of the input image. We formulate image downscaling as an optimization problem where the difference between the input and output images is measured using a widely adopted perceptual image quality metric. The downscaled images retain perceptually important features and details, resulting in an accurate and spatio-temporally consistent representation of the high resolution input. We derive the solution of the optimization problem in closed-form, which leads to a simple, efficient and parallelizable implementation with sums and convolutions. The algorithm has running times similar to linear filtering and is orders of magnitude faster than the state-of-the-art for image downscaling. We validate the effectiveness of the technique with extensive tests on many images, video, and by performing a user study, which indicates a clear preference for the results of the new algorithm.

Downloads

Download Paper
[PDF]
Download Video
[Video]