Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
IEICE Technical Committee Submission System
Conference Paper's Information		 Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2026-02-19 17:00
A Two-Stage Lightweight Framework for Learned Image Compression via Distillation and Quantization
Zhen Zhou, Heming Sun (YNU) ITS2025-60 IE2025-75
Abstract (in Japanese) (See Japanese page) 
(in English) In recent years, learned image compression (LIC) showed a powerful performance in the image compression task compared to the widely used traditional codecs such as JPEG. Furthermore, the modern LIC model structure that based on the variational autoencoder (VAE), has an additional hyper-path to capture detailed information in the latent space. The hyper-path gives the network a better probabilistic modeling ability for data encoding. However, although it was designed to collect sub-information, the hyper-path could consume 40% of the whole network's parameters. We consider this part of consumption to be unnecessary and unbalanced. To solve this problem, we proposed a two-step lightweight training framework that is based on knowledge distillation and quantization, and it can compress the hyper-path in both the structure and numerical field. Although knowledge distillation and quantization are rapidly adopted in network compression, there are few research that has combined them, especially for the hyper-path of the LIC model. The results show that our proposal successfully decreased the hyper-path's parameters and storage in 59.8% and 97.5% separately. Meanwhile, the accuracy only reduced 7.2% on average. As a conclusion, our research demonstrate an lightweight and high-quality compression framework for the hyper-path.
Keyword (in Japanese) (See Japanese page) 
(in English) learned image compression / / / / / / /  
Reference Info. IEICE Tech. Rep., vol. 125, no. 356, IE2025-75, pp. 119-123, Feb. 2026.
Paper # IE2025-75 
Date of Issue 2026-02-12 (ITS, IE) 
ISSN Online edition: ISSN 2432-6380
Copyright
and
reproduction		 All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF ITS2025-60 IE2025-75

Conference Information
Committee IE ITS ITE-MMS ITE-ME ITE-AIT ITE-SIP  
Conference Date 2026-02-19 - 2026-02-20 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To IE 
Conference Code 2026-02-IE-ITS-MMS-ME-AIT-SIP 
Language English 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) A Two-Stage Lightweight Framework for Learned Image Compression via Distillation and Quantization 
Sub Title (in English)  
Keyword(1) learned image compression  
Keyword(2)  
Keyword(3)  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Zhen Zhou  
1st Author's Affiliation Yokohama National University (YNU)
2nd Author's Name Heming Sun  
2nd Author's Affiliation Yokohama National University (YNU)
3rd Author's Name  
3rd Author's Affiliation ()
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2026-02-19 17:00:00 
Presentation Time 15 minutes 
Registration for IE 
Paper # ITS2025-60, IE2025-75 
Volume (vol) vol.125 
Number (no) no.355(ITS), no.356(IE) 
Page pp.119-123 
#Pages 5 
Date of Issue 2026-02-12 (ITS, IE) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan