A Learning-Based Approach to Reduce JPEG Artifacts in Image Matting
Authors:
- Inchang ChoiICInchang Choi
- Sunyeong KimSKSunyeong Kim
- Michael S. BrownMBMichael S. Brown
- Yu-Wing TaiYTYu-Wing Tai
- 1 December 2013
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15505499,p. 2880-2887
Abstract
Single image matting techniques assume high-quality input images. The vast majority of images on the web and in personal photo collections are encoded using JPEG compression. JPEG images exhibit quantization artifacts that adversely affect the performance of matting algorithms. To address this situation, we propose a learning-based post-processing method to improve the alpha mattes extracted from JPEG images. Our approach learns a set of sparse dictionaries from training examples that are used to transfer details from high-quality alpha mattes to alpha mattes corrupted by JPEG compression. Three different dictionaries are defined to accommodate different object structure (long hair, short hair, and sharp boundaries). A back-projection criteria combined within an MRF framework is used to automatically select the best dictionary to apply on the object's local boundary. We demonstrate that our method can produces superior results over existing state-of-the-art matting algorithms on a variety of inputs and compression levels.Keywords
This publication has 21 references indexed in Scilit:
- Natural Matting for Degraded PicturesIEEE Transactions on Image Processing, 2011
- Motion Regularization for Matting Motion Blurred ObjectsIEEE Transactions on Pattern Analysis and Machine Intelligence, 2011
- Shared Sampling for Real‐Time Alpha MattingComputer Graphics Forum, 2010
- Image and Video Matting: A SurveyFoundations and Trends in Computer Graphics and Vision, 2008
- Optimized Color Sampling for Robust Matting2007
- Learning Sparse Overcomplete Codes for ImagesJournal of Signal Processing Systems, 2006
- Poisson mattingACM Transactions on Graphics, 2004
- Comparison of graph cuts with belief propagation for stereo, using identical MRF parameters2003
- The JPEG still picture compression standardCommunications of the ACM, 1991
- Compositing digital imagesACM SIGGRAPH Computer Graphics, 1984