Peering into The Sketch: Ultra-Low Bitrate Face Compression for Joint Human and Machine Perception
Open Access
- 26 October 2023
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 2564-2572
Abstract
We propose a novel face compression framework that leverages the external priors for joint human and machine perception under ultra-low bitrate scenarios. The proposed framework leverages the semantic richness of face images by representing the faces into sketches and thumbnails, resulting in improved bitrate utility for both human and machine vision. At the decoder side, the framework introduces a two-stage generative reconstruction, which faithfully enhances the reconstructed image via semi-parametric modeling and retrieved guidance from the external database. In particular, this coarse-to-fine strategy also results in improved identity consistency and analysis performance of the reconstructed image. Extensive evaluations of the proposed method have been conducted on the public face dataset by comparing it with end-to-end image compression techniques as well as traditional image compression standards. The experimental results demonstrate the effectiveness of the proposed method via superior perceptual and analytical performance under ultra-low bitrate conditions.Keywords
Funding Information
- The Science Technology and Innovation Committee of Shenzhen Municipality (2021Szvup128)
- The National Natural Science Foundation of China (62022002)
- the Hong Kong Research Grants Council General Research Fund (GRF) (9042957)
This publication has 31 references indexed in Scilit:
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric2018
- Deformable Convolutional Networks2017
- Joint Face Detection and Alignment Using Multitask Cascaded Convolutional NetworksIEEE Signal Processing Letters, 2016
- Learning to simplifyACM Transactions on Graphics, 2016
- End-to-end optimization of nonlinear transform codes for perceptual quality2016
- Deep Learning Face Attributes in the Wild2015
- Holistically-Nested Edge Detection2015
- Landmark Image Super-Resolution by Retrieving Web ImagesIEEE Transactions on Image Processing, 2013
- JPEG2000: Image Compression Fundamentals, Standards and PracticeJournal of Electronic Imaging, 2002
- The JPEG still picture compression standardCommunications of the ACM, 1991