Image compression with xvc

Jonatan Samuelsson, Per Hermansson; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 2595-2597

Abstract


This paper describes xvc - a format for efficient compression of visual data - originally developed for compression of video sequences, but in the context of this paper applied to still images. The xvc codec is a block based lossy codec using a traditional approach for prediction, residual representation and entropy coding. There are no elements of Machine Learning or Artificial Neural Networks in the xvc encoder or decoder. The xvc codec offers support for tuning towards PSNR or perceptual quality. The images submitted for the CLIC challenge and the descriptions included in this paper are based on the perceptually tuned setting.

Related Material


[pdf]
[bibtex]
@InProceedings{Samuelsson_2018_CVPR_Workshops,
author = {Samuelsson, Jonatan and Hermansson, Per},
title = {Image compression with xvc},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2018}
}