Jyrki Alakuijala, Robert Obryk, Zoltan Szabadka, Jan Wassenberg
2017 · arXiv
We report on pairwise comparisons by human raters of JPEG images from libjpeg and our new Guetzli encoder. Although both files are size-matched, 75% of ratings are in favor of Guetzli. This implies the Butteraugli psychovisual image similarity metric which guides Guetzli is reasonably close to human perception at high quality levels. We provide access to the raw ratings and source images for further analysis and study.
Metadata: arXiv. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
3 selected excerpts · 3 total mentions
There are glaring optimization issues. Over 6MB of those 10MB is just one bloated photo [1]. It can be compressed down to 1-2MB. Moreover, if it was a proper responsive image, it would not be larger than 100-200 KB for most readers because a good JPEG compressor can pull 2 bits per pixel [2]. Ironically, image optimization apps don't get that much coverage. [1] https://techcrunch.com/wp-content/uploads/2018/05/vikbajaj20... [2] https://arxiv.org/pdf/1703.04416.pdf
I find this article a bit poor. It basically gives a few data points about the resulting image size for a few quality settings from an unknown source image already compressed in jpeg with unknown quality setting. No example of perceptual quality degradation or screenshots (which would have been useful for a perceptual codec) besides him finding no and another person find very little difference. No comparison to a standard jpeg compressor for the same image / quality. I'm curious : how does guetzli that compare to libjpeg ? libjpeg-turbo ? whatever photoshop is using ?) esp in size for the same quality on your image ? I find the original blog post is better (having links…
and actual screenshots): https://research.googleblog.com/2017/03/announcing-guetzli-n... the Arxiv paper is much better controlled on how the experiment went, but has few/no figures besides the fact that it's better: https://arxiv.org/pdf/1703.04416.pdf EDIT : I find this article more complete : https://www.34sp.com/blog/speed-testing-googles-guetzli-jpeg...
We did an experiment with comparisons of Guetzli and (slightly larger) libjpeg output: https://arxiv.org/abs/1703.04416 Turns out that 75% of the 614 ratings are in favor of the Guetzli version.
More from the same topics.
Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan +11
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
A. Pluchino. A. E. Biondo, A. Rapisarda
The largely dominant meritocratic paradigm of highly competitive Western cultures is rooted on the belief that success is due mainly, if not exclusively, to personal qualities such as talent, intelligence, skills, efforts or risk taking.
David Manheim, Scott Garrabrant
There are several distinct failure modes for overoptimization of systems on the basis of metrics.
Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai +3
Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools.