Alpha Matting Evaluation Website

Image Matting Evaluation•Datasets•Code•Submit

How to use this page?

Please be patient until all images have loaded completely.
Move the mouse over the numbers in the table to see the corresponding images.
Drag the red rectangle in the leftmost image to change the location of the zoom.
Press and hold any key to temporarily deactivate the links.

Image matting evaluation results Competition: Low resolution High resolution
Error type: SAD MSE Gradient Connectivity

Gradient error overall
avg.
small avg.
large avg.
user Troll
(Strongly Transparent)
Input Doll
(Strongly Transparent)
Input Donkey
(Medium Transparent)
Input Elephant
(Medium Transparent)
Input Plant
(Little Transparent)
Input Pineapple
(Little Transparent)
Input Plastic bag
(Highly Transparent)
Input Net
(Highly Transparent)
Input

rank rank rank rank small large user small large user small large user small large user small large user small large user small large user small large user

Shared Matting 2.9 2.8 3.3 2.8 0.2 1 0.3 3 0.2 2 0.3 8 0.6 8 0.5 8 0.4 1 0.4 1 0.4 1 0.4 2 0.8 5 0.5 1 1.6 1 2 2 2.3 1 1.4 1 1.9 1 1.8 1 1.5 5 1.5 3 1.1 5 0.6 3 0.6 3 0.4 3

Segmentation-based matting 3 3.8 2.8 2.5 0.2 2 0.3 1 0.2 1 0.2 2 0.2 1 0.3 2 0.4 3 0.5 4 0.4 2 0.4 1 0.6 1 0.5 2 2.1 5 1.7 1 2.5 2 1.9 4 2.2 3 2.5 4 2 11 1.7 9 1.2 6 0.5 2 0.5 2 0.4 1

Improved color matting 3.2 3.6 3 2.9 0.2 3 0.3 2 0.2 3 0.2 5 0.3 4 0.3 3 0.4 2 0.5 3 0.4 3 0.5 6 0.7 4 0.5 5 2.5 7 2.5 5 2.7 3 1.9 3 2.5 4 2.4 3 1.4 2 1.4 1 0.9 1 0.4 1 0.4 1 0.4 2

Closed-Form Matting 5.8 5.9 5.5 6.1 0.3 5 0.3 5 0.3 5 0.2 1 0.3 3 0.3 5 0.4 5 0.5 6 0.4 4 0.4 4 0.6 3 0.6 6 3.3 10 3.6 9 4.6 9 2.9 10 3.5 8 3.7 10 1.5 6 1.5 4 0.9 2 0.7 6 0.8 6 1.2 8

Learning Based Matting 5.9 5.6 5.8 6.3 0.3 8 0.3 6 0.3 7 0.2 3 0.3 2 0.3 4 0.4 7 0.5 7 0.4 6 0.4 3 0.6 2 0.5 4 1.9 3 2.5 4 3.2 6 3.1 11 3.6 10 3.9 11 1.4 1 1.6 6 1 3 0.9 9 1.1 9 1.2 9

Robust Matting 6 5.3 6 6.9 0.3 6 0.4 8 0.3 8 0.5 9 0.9 9 0.6 9 0.4 6 0.5 2 0.5 8 0.5 7 0.8 6 0.7 8 1.8 2 2.6 6 3.1 5 2 5 3.1 6 2.7 5 1.4 3 1.6 7 1.2 8 0.6 4 0.6 4 0.5 4

High-res matting 6.5 5.9 7.9 5.9 0.2 4 0.3 4 0.3 4 0.2 6 0.5 7 0.4 7 0.4 4 0.5 5 0.4 5 0.4 5 1.3 10 0.5 3 2.3 6 2.9 7 3.7 8 2.6 7 3.8 11 3.3 8 1.8 10 1.8 11 1.2 7 0.6 5 1.1 8 0.6 5

Shared Matting (Real Time) 6.8 7.1 6.4 6.9 0.3 9 0.5 10 0.4 10 0.6 10 0.9 10 0.7 10 0.4 8 0.5 8 0.4 7 0.6 9 1.1 8 0.6 7 2 4 2.5 3 2.9 4 1.7 2 2.1 2 2 2 1.6 8 1.6 5 1.3 9 0.8 7 0.8 5 0.7 6

Large Kernel Matting 6.8 7.1 6.9 6.5 0.3 7 0.3 7 0.3 6 0.2 4 0.3 5 0.2 1 0.5 9 0.6 10 0.5 9 0.5 8 1.3 9 0.7 9 2.9 9 3.3 8 3.7 7 2.6 8 3.3 7 3.4 9 1.4 4 1.4 2 1.1 4 0.8 8 0.8 7 0.8 7

Iterative BP Matting 10.3 9.8 10.5 10.8 0.5 11 0.5 11 0.6 11 0.7 11 1.2 12 1.2 12 0.6 11 0.8 11 0.6 11 0.7 11 1.5 11 1.1 11 2.7 8 3.6 10 4.9 11 2.9 9 3.6 9 3.2 7 1.6 7 1.8 10 2.9 13 1.1 10 1.3 10 1.2 10

Random Walk Matting 10.5 10.9 9.9 10.6 0.4 10 0.5 9 0.4 9 0.3 7 0.4 6 0.3 6 0.5 10 0.6 9 0.5 10 0.7 10 1 7 0.8 10 4.8 14 6.4 14 6.7 14 4.4 14 5 14 4.7 14 1.7 9 1.6 8 1.3 10 1.7 13 1.9 12 1.7 12

Geodesic Matting 11.8 12.3 11.6 11.6 1.4 14 3.2 15 1.8 14 1.2 12 1.1 11 0.9 11 1.2 14 1.5 15 1 14 1.9 14 2.1 13 1.7 14 4.1 13 4.6 11 4.6 10 2.2 6 3 5 2.9 6 3.5 13 3.3 12 2 11 1.3 12 1.4 11 1.8 13

Bayesian Matting 12.7 12.4 13.1 12.6 1.3 13 1.5 13 1.2 13 1.8 13 2 13 1.6 13 0.9 13 1.1 13 1 13 1.3 13 2.5 14 1.2 13 3.6 11 5.8 13 6.6 13 4 13 4.7 13 4.1 13 3.1 12 3.6 13 2.8 12 1.2 11 2 13 1.3 11

Easy Matting 12.8 12.8 12.9 12.8 1.1 12 1.3 12 1 12 2.5 14 3 14 2.8 14 0.7 12 0.8 12 0.7 12 1.1 12 2.1 12 1.1 12 4.1 12 4.6 12 6.3 12 3.7 12 4.1 12 4.1 12 3.6 14 4.5 15 3.4 14 2.9 14 3.1 14 3.3 14

Poisson Matting 14.9 15 14.6 15 2.3 15 2 14 2.6 15 3.4 15 3.6 15 3.4 15 1.5 15 1.4 14 1.3 15 3.1 15 3 15 3.1 15 6.6 15 7.9 15 9.9 15 6.7 15 7.4 15 6.6 15 3.8 15 4.2 14 4.5 15 4.1 15 5.3 15 4.3 15

Move the mouse over the numbers in the table to see the corresponding images.Click to compare with the ground truth. Press a key to deactivate the links (to better use the zoom).

Gradient error	overall	avg. small	avg. large	avg. user	Troll (Strongly Transparent) Input	Doll (Strongly Transparent) Input	Donkey (Medium Transparent) Input	Elephant (Medium Transparent) Input	Plant (Little Transparent) Input	Pineapple (Little Transparent) Input	Plastic bag (Highly Transparent) Input	Net (Highly Transparent) Input
Shared Matting	2.9	2.8	3.3	2.8	0.2 1	0.3 3	0.2 2	0.3 8	0.6 8	0.5 8	0.4 1	0.4 1	0.4 1	0.4 2	0.8 5	0.5 1	1.6 1	2 2	2.3 1	1.4 1	1.9 1	1.8 1	1.5 5	1.5 3	1.1 5	0.6 3	0.6 3	0.4 3
Segmentation-based matting	3	3.8	2.8	2.5	0.2 2	0.3 1	0.2 1	0.2 2	0.2 1	0.3 2	0.4 3	0.5 4	0.4 2	0.4 1	0.6 1	0.5 2	2.1 5	1.7 1	2.5 2	1.9 4	2.2 3	2.5 4	2 11	1.7 9	1.2 6	0.5 2	0.5 2	0.4 1
Improved color matting	3.2	3.6	3	2.9	0.2 3	0.3 2	0.2 3	0.2 5	0.3 4	0.3 3	0.4 2	0.5 3	0.4 3	0.5 6	0.7 4	0.5 5	2.5 7	2.5 5	2.7 3	1.9 3	2.5 4	2.4 3	1.4 2	1.4 1	0.9 1	0.4 1	0.4 1	0.4 2
Closed-Form Matting	5.8	5.9	5.5	6.1	0.3 5	0.3 5	0.3 5	0.2 1	0.3 3	0.3 5	0.4 5	0.5 6	0.4 4	0.4 4	0.6 3	0.6 6	3.3 10	3.6 9	4.6 9	2.9 10	3.5 8	3.7 10	1.5 6	1.5 4	0.9 2	0.7 6	0.8 6	1.2 8
Learning Based Matting	5.9	5.6	5.8	6.3	0.3 8	0.3 6	0.3 7	0.2 3	0.3 2	0.3 4	0.4 7	0.5 7	0.4 6	0.4 3	0.6 2	0.5 4	1.9 3	2.5 4	3.2 6	3.1 11	3.6 10	3.9 11	1.4 1	1.6 6	1 3	0.9 9	1.1 9	1.2 9
Robust Matting	6	5.3	6	6.9	0.3 6	0.4 8	0.3 8	0.5 9	0.9 9	0.6 9	0.4 6	0.5 2	0.5 8	0.5 7	0.8 6	0.7 8	1.8 2	2.6 6	3.1 5	2 5	3.1 6	2.7 5	1.4 3	1.6 7	1.2 8	0.6 4	0.6 4	0.5 4
High-res matting	6.5	5.9	7.9	5.9	0.2 4	0.3 4	0.3 4	0.2 6	0.5 7	0.4 7	0.4 4	0.5 5	0.4 5	0.4 5	1.3 10	0.5 3	2.3 6	2.9 7	3.7 8	2.6 7	3.8 11	3.3 8	1.8 10	1.8 11	1.2 7	0.6 5	1.1 8	0.6 5
Shared Matting (Real Time)	6.8	7.1	6.4	6.9	0.3 9	0.5 10	0.4 10	0.6 10	0.9 10	0.7 10	0.4 8	0.5 8	0.4 7	0.6 9	1.1 8	0.6 7	2 4	2.5 3	2.9 4	1.7 2	2.1 2	2 2	1.6 8	1.6 5	1.3 9	0.8 7	0.8 5	0.7 6
Large Kernel Matting	6.8	7.1	6.9	6.5	0.3 7	0.3 7	0.3 6	0.2 4	0.3 5	0.2 1	0.5 9	0.6 10	0.5 9	0.5 8	1.3 9	0.7 9	2.9 9	3.3 8	3.7 7	2.6 8	3.3 7	3.4 9	1.4 4	1.4 2	1.1 4	0.8 8	0.8 7	0.8 7
Iterative BP Matting	10.3	9.8	10.5	10.8	0.5 11	0.5 11	0.6 11	0.7 11	1.2 12	1.2 12	0.6 11	0.8 11	0.6 11	0.7 11	1.5 11	1.1 11	2.7 8	3.6 10	4.9 11	2.9 9	3.6 9	3.2 7	1.6 7	1.8 10	2.9 13	1.1 10	1.3 10	1.2 10
Random Walk Matting	10.5	10.9	9.9	10.6	0.4 10	0.5 9	0.4 9	0.3 7	0.4 6	0.3 6	0.5 10	0.6 9	0.5 10	0.7 10	1 7	0.8 10	4.8 14	6.4 14	6.7 14	4.4 14	5 14	4.7 14	1.7 9	1.6 8	1.3 10	1.7 13	1.9 12	1.7 12
Geodesic Matting	11.8	12.3	11.6	11.6	1.4 14	3.2 15	1.8 14	1.2 12	1.1 11	0.9 11	1.2 14	1.5 15	1 14	1.9 14	2.1 13	1.7 14	4.1 13	4.6 11	4.6 10	2.2 6	3 5	2.9 6	3.5 13	3.3 12	2 11	1.3 12	1.4 11	1.8 13
Bayesian Matting	12.7	12.4	13.1	12.6	1.3 13	1.5 13	1.2 13	1.8 13	2 13	1.6 13	0.9 13	1.1 13	1 13	1.3 13	2.5 14	1.2 13	3.6 11	5.8 13	6.6 13	4 13	4.7 13	4.1 13	3.1 12	3.6 13	2.8 12	1.2 11	2 13	1.3 11
Easy Matting	12.8	12.8	12.9	12.8	1.1 12	1.3 12	1 12	2.5 14	3 14	2.8 14	0.7 12	0.8 12	0.7 12	1.1 12	2.1 12	1.1 12	4.1 12	4.6 12	6.3 12	3.7 12	4.1 12	4.1 12	3.6 14	4.5 15	3.4 14	2.9 14	3.1 14	3.3 14
Poisson Matting	14.9	15	14.6	15	2.3 15	2 14	2.6 15	3.4 15	3.6 15	3.4 15	1.5 15	1.4 14	1.3 15	3.1 15	3 15	3.1 15	6.6 15	7.9 15	9.9 15	6.7 15	7.4 15	6.6 15	3.8 15	4.2 14	4.5 15	4.1 15	5.3 15	4.3 15

Troll - Input image
Drag the window to change the zoom.

References

Method	Reference and notes	Implementation details
Closed-Form Matting	A. Levin, D. Lischinski, Y. Weiss, A Closed Form Solution to Natural Image Matting, CVPR, 2006	Matlab implementation on a Intel Core2 Quad with 2.4 GHZ
Bayesian Matting	Y.Y. Chuang, B. Curless, D. Salesin, R. Szeliski, A Bayesian Approach to Digital Matting, CVPR, 2001	C++ implementation on a Intel Core2 Quad with 2.4 GHZ
Poisson Matting	J. Sun, J. Jia, C.K. Tang, H.Y. Shum, Poisson matting, SIGGRAPH, 2004	Matlab implementation on a Intel Core2 Quad with 2.4 GHZ
Easy Matting	Y. Guan, W. Cheny, X. Liang, Z. Ding, Q. Peng, Easy Matting: A Stroke Based Approach for Continuous Image Matting, Eurographics, 2006	C++ implementation on a Intel Core2 Quad with 2.4 GHZ
Random Walk Matting	L. Grady, T. Schiwietz, S. Aharon, Random Walks For Interactive Alpha-Matting, VIIP, 2005	Matlab/C++ implementation on a Intel Core2 Quad with 2.4 GHZ
Robust Matting	J. Wang, M. Cohen, Optimized Color Sampling for Robust Matting, CVPR, 2007	C++ implementation on a Intel Core2 Quad with 2.4 GHZ
Geodesic Matting	Xue Bai, Guillermo Sapiro, A geodesic framework for fast interactive image and video segmentation and matting, ICCV 2007	C++ implementation on a Intel Core2 Duo with 2.53 GHZ
Iterative BP Matting	Jue Wang, Michael Cohen, Jue Wang and Michael F. Cohen. An iterative optimization approach for unified image segmentation and matting. ICCV 2005.	c++ implementation on a Intel Core2 Quad with 3 GHZ
Improved color matting	C. Rhemann, C. Rother, M. Gelautz, Improving Color Modeling for Alpha Matting. BMVC, 2008	Matlab implementation on a Intel Core2 Duo with 2.4 GHZ
High-res matting	C. Rhemann, C. Rother, A. Rav-Acha, M. Gelautz, T. Sharp, High ResolutionMatting via Interactive Trimap Segmentation. CVPR, 2008	Matlab/C++ implementation on a Intel Core2 Duo with 2.4 GHZ
Large Kernel Matting	Kaiming He, Jian Sun, and Xiaoou Tang, Fast Matting using Large Kernel Matting Laplacian Matrices, CVPR 2010	C++ implementation on a Intel Core Duo with 2 GHZ
Segmentation-based matting	Christoph Rhemann, Carsten Rother, Pushmeet Kohli, Margrit Gelautz, A Spatially Varying PSF-based Prior for Alpha Matting, CVPR 2010	Matlab/C++ implementation on a Intel Core2 Quad with 2.39 GHZ
Shared Matting	Eduardo S. L. Gastal and Manuel M. Oliveira, Shared Sampling for Real-Time Alpha Matting, Eurographics, 2010	C++/GLSL implementation on a Core 2 Quad with 2.8 GHZ
Shared Matting (Real Time)	Eduardo S. L. Gastal and Manuel M. Oliveira, Shared Sampling for Real-Time Alpha Matting, Eurographics, 2010	C++/GLSL implementation on a Core 2 Quad with 2.8 GHZ
Learning Based Matting	Yuanjie Zheng, Chandra Kambhamettu, Yuanjie Zheng, Chandra Kambhamettu. Learning Based Digital Matting. ICCV 2009	Matlab/C++ implementation on a Intel Core2 Duo with 2.53 GHZ