A new end-to-end image compression system based on convolutional neural networks

In this paper, two new end-to-end image compression architectures based on convolutional neural networks are presented. The proposed networks employ 2D wavelet decomposition as a preprocessing step before training and extract features for compression from wavelet coefficients. Training is performed end-to-end and multiple models operating at di↵erent rate points are generated by using a regularizer in the loss function. Results show that the proposed methods outperform JPEG compression, reduce blocking and blurring artifacts, and preserve more details in the images especially at low bitrates.


Editor(s):
Andrew G. Tescher, Touradj Ebrahimi
Published in:
Applications of Digital Image Processing XLII, 22
Presented at:
SPIE Optical Engineering + Applications, San Diego, California, US, 11 - 15 August 2019
Year:
Sep 06 2019
Other identifiers:
Laboratories:




 Record created 2019-09-12, last modified 2019-09-13

Fulltext:
Download fulltext
PDF

Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)