arrow
返回

Semantic Image Matting: General and Specific Semantics

delete2023-10-12
delete1
PRE
AI
Y
Yanan Sun *
C
Chi–Keung Tang
Y
Yu‐Wing Tai
DOI:10.1007/s11263-023-01907-6delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
Although conventional matting formulation can separate foreground from background in fractional occupancy which can be caused by highly transparent objects, complex foreground (e.g., net or tree), and objects containing very fine details (e.g., hairs), no previous work has attempted to reason the underlying causes of matting due to various foreground semantics in general. We show how to obtain better alpha mattes by incorporating into our framework semantic classification of matting regions. Specifically, we consider and learn 20 classes of general matting patterns, and propose to extend the conventional trimap to semantic trimap. The proposed semantic trimap can be obtained automatically through patch structure analysis within trimap regions. Meanwhile, we learn a multi-class discriminator to regularize the alpha prediction at semantic level, and content-sensitive weights to balance different regularization losses. Experiments on multiple benchmarks show that our method outperforms other methods benefit from such general alpha semantics and has achieved the most competitive state-of-the-art performance. We further explore the effectiveness of our method on specific semantics by specializing our method into human matting and transparent object matting. Experimental results on specific semantics demonstrate alpha matte semantic information can boost performance for not only general matting but also class-specific matting. Finally, we contribute a large-scale Semantic Image Matting Dataset constructed with careful consideration of data balancing across different semantic classes. Code and dataset are available in https://github.com/nowsyn/SIM.
Keyword:
Image matting
Semantics
Classification
Class-specific matting

期刊

International Journal of Computer Vision 封面图
International Journal of Computer Vision
IF:
9.3
论文数:
3.9K
被引数:
2.8W

机构

D
Dartmouth College
学者数:
1.5W
论文数: 1.4W
被引数: 1.8W
引用论文

引用论文

How, When, and Where do Football Players get Injured?: a Descriptive Epidemiological Study on Male Professional Football Players in Spain for Four Seasons
err2017-10-01
err0
errOAAI
errPedro Gomez-Piqueras; Alberto Nájera López; Jesus Gonzalez-Rubio; Enrique Arribas; María del Pilar Sainz de Baranda Andújar
err分享
err收藏
Structural and magnetic engineering of (Nd, Pr, Dy, Tb)–Fe–B sintered magnets with Tb3Co0.6Cu0.4H composition in the powder mixture
err2020-03-01
err0
PREAI
errKaterina Skotnicova; Gennady S. Burkhanov; Natalia B. Kolchugina; Miroslav Kursa; Tomas Cegan; Alexander A. Lukin; Ondrej Zivotsky; Pavel A. Prokofev; Jan Jurica; Ying Li
err分享
err收藏
Matting and Compositing of Transparent and Refractive Objects
err2011-02-02
err24
PREAI
errYeung, Sai-Kit; Tang, Chi-Keung; Brown, Michael S.; Kang, Sing Bing
err分享
err收藏
err分享
err收藏
Vitamin D deficiency and exercise-induced laryngospasm in young competitive rowers
err2016-07-01
err0
PREAI
errEnrico Heffler; Matteo Bonini; Luisa Brussino; Paolo Solidoro; Giuseppe Guida; Monica Boita; Giuliana Nicolosi; Caterina Bucca
err分享
err收藏
Clinical and radiological cervical spine evaluation in retired professional rugby players
err2015-11-01
err0
errOAAI
errDavid Brauge; Cyrille Delpierre; Philippe Adam; Jean Christophe Sol; Pierre Bernard; Franck-Emmanuel Roux
err分享
err收藏
学者 查看更多内容