Meta tags:
description=VideoMatting project is the first public objective benchmark of video matting methods. It is based on a high-quality ground truth alpha sequences collected using both capturing in front of a green plate and stop-motion (sequential photography).We believe that our work will help ranking existing methods and challenge authors of new methods.;

Headings (most frequently used words):

evaluation, videomatting, overview, references, what, new, introduction, datasets, methodology, public, sequences, participate, cite, us, rating, integral, plots, subjective, comparison, multidimensional, analysis, chroma, keying, stop, motion, graphics, media, lab, lomonosov, moscow, state, university,

Text of the page (most frequently used words):
matting (91), the (66), and (59), tion (40), ing (37), for (32), ting (27), 2016 (27), com (26), mat (26), ground (21), 2008 (21), 2020 (20), 2012 (19), sampling (19), pro (15), based (15), with (14), doi (14), video (14), our (14), sion (13), 2017 (13), form (13), flow (12), puter (12), pat (12), 2013 (12), sparse (12), #evaluation (12), method (11), tern (11), self (11), deep (11), edge (11), 2009 (11), knn (11), information (11), trans (10), age (10), con (10), tions (10), pages (10), 2007 (10), closed (10), 2011 (10), shared (10), 2010 (10), bayesian (10), 2001 (10), truth (10), fba (10), background (10), adaptive (10), refine (10), are (9), pha (9), ject (9), sequences (9), spectral (9), comprehensive (9), nonlocal (9), learning (9), robust (9), back (8), ter (8), recog (8), cvpr (8), per (8), meth (8), ods (8), screen (8), stop (8), published (8), results (8), code (7), fer (7), par (7), each (7), use (7), dmitriy (6), set (6), data (6), sam (6), page (6), subjective (6), quences (6), cap (6), new (6), jue (5), wang (5), nat (5), ural (5), can (5), color (5), videos (5), these (5), quality (5), that (5), low (5), your (5), 110 (5), will (5), vel (5), pub (5), sults (5), met (5), rics (5), mattes (5), plots (4), fore (4), pixel (4), rank (4), ence (4), analy (4), sis (4), hen (4), ent (4), gence (4), pling (4), tal (4), bench (4), mark (4), chine (4), cal (4), less (4), from (4), then (4), study (4), participants (4), were (4), trimap (4), 104 (4), 122 (4), 112 (4), available (4), benchmark (4), cite (4), this (4), parency (4), toy (4), front (4), cam (4), era (4), motion (4), chroma (4), green (4), added (4), mikhail (3), yury (3), alexey (3), anony (3), mous (3), sub (3), mis (3), cess (3), tive (3), chen (3), tional (3), web (3), bradley (3), terry (3), adobe (3), dig (3), tel (3), tpami (3), analysis (3), comparison (3), methods (3), using (3), platform (3), foreground (3), them (3), better (3), they (3), used (3), you (3), woods (3), juneau (3), artem (3), vitaliy (3), slava (3), snow (3), flowers (3), concert (3), rain (3), city (3), 108 (3), 118 (3), 115 (3), 157 (3), 113 (3), rating (3), eval (3), test (3), work (3), alpha (3), ers (3), lish (3), fol (3), participate (3), maps (3), public (3), cor (3), tor (3), not (3), son (3), look (3), light (3), flec (3), able (3), ages (3), mon (3), tures (3), key (3), jects (3), trimaps (3), was (3), now (3), videomatting (2), erofeev (2), gitman (2), vatolin (2), fedorov (2), tract (2), ieee (2), pair (2), crowd (2), acm (2), min (2), res (2), brian (2), price (2), tool (2), fects (2), lev (2), dem (2), tolin (2), shahrian (2), jan (2), prov (2), pre (2), ture (2), cep (2), ally (2), vated (2), levin (2), lischin (2), ski (2), 242 (2), non (2), graph (2), 2015 (2), british (2), bmvc (2), multidimensional (2), containing (2), subjectify (2), pairwise (2), including (2), control (2), comparisons (2), answers (2), since (2), one (2), integral (2), sure (2), note (2), composite (2), source (2), 106 (2), 128 (2), 130 (2), 127 (2), 123 (2), 129 (2), 148 (2), 120 (2), 117 (2), 121 (2), 102 (2), 103 (2), 169 (2), 138 (2), 107 (2), 159 (2), 125 (2), 109 (2), 135 (2), 144 (2), year (2), frame (2), 5244 (2), org (2), any (2), ques (2), about (2), ask (2), cases (2), scores (2), oper (2), also (2), tic (2), sim (2), ply (2), steps (2), here (2), their (2), notes (2), els (2), spond (2), fair (2), have (2), tailed (2), scrip (2), qual (2), ity (2), given (2), dif (2), methodology (2), end (2), changes (2), than (2), noise (2), spe (2), cial (2), tice (2), amount (2), solid (2), same (2), dure (2), jec (2), two (2), step (2), capture (2), over (2), backgrounds (2), cin (2), five (2), get (2), ance (2), keying (2), mor (2), pho (2), log (2), more (2), mov (2), tured (2), datasets (2), view (2), introduction (2), overview (2), levels (2), paper (2), natural (2), see (2), what (2), 2014, powered, highcharts, project, contributors, graphics, media, lab, lomonosov, moscow, state, university, eccv, 6826, novel, clean, giz, soy, tunc, ozan, din, marc, polle, feys, sign, fec, wise, gre, sourced, ceed, ings, sixth, search, ralph, lan, mil, ton, plete, block, signs, paired, isons, bio, metrika, 324, 345, 1952, guangy, cao, jian, wei, aowu, zhiqiang, patch, adap, high, sual, ning, scott, thomas, huang, fine, http, www, prod, ucts, html, kara, aykut, erkut, ver, yuan, jie, zheng, kamb, hamettu, learn, 889, 896, iccv, michael, mized, bust, karen, symonyan, sergey, ishin, popov, fast, via, clas, 349, 352, icip, sive, sets, 636, 643, weighted, tex, ple, lec, 718, 725, christoph, rhe, mann, carsten, rother, margrit, gelautz, push, meet, kohli, pamela, rott, line, 1826, 1833, anat, dani, yair, weiss, 228, rav, acha, spec, tral, 1699, 1712, philip, lee, ying, 2193, 2200, uardo, gastal, manuel, oliveira, real, time, 575, 584, ics, rum, ero, feev, git, man, dorov, pdf, yung, chuang, cur, david, salesin, richard, szeliski, proach, vol, ume, 264, 271, qifeng, dingzeyu, chi, ung, tang, 2175, 2188, references, carried, out, applied, dataset, uploaded, extracted, objects, hired, showed, fashion, asked, choose, visual, indicate, approximately, equal, participant, compared, pairs, hidden, between, rejected, failed, least, question, total, 10556, 406, collected, models, convert, ranks, report, generated, shown, below, make, latest, version, browser, recommend, chromium, browsers, fbam, mwbe, ifm, spsm, spm, shm, nlm, 183, 119, 155, 114, 151, 111, 131, 193, 218, 116, 181, 241, 143, 179, 170, 100, 101, 188, 162, 214, 176, 154, 182, 210, 105, 171, 243, 134, 149, 140, wide, medium, narrow, size, inproceedings, erofeev2015, title, perceptually, motivated, author, month, september, articleno, numpages, booktitle, proceedings, machine, vision, conference, publisher, bmva, press, isbn, 901725, url, https, tact, email, sug, ges, fill, provide, upload, file, sharing, service, kindly, maintain, doesn, explicitly, produce, images, skip, uploading, case, generate, proposed, naming, directory, structure, conventions, apply, download, vite, ate, mit, ted, port, where, specif, cally, grants, site, blind, viewed, pers, pate, alex, castle, train, pur, poses, three, searchers, wel, come, but, num, ber, pix, val, ues, der, sid, ingly, vec, rithm, puted, worth, aware, give, van, tage, lar, quence, cludes, both, pare, racy, tem, ral, herency, into, rected, global, caused, bulb, flick, thus, nally, tain, level, trac, paid, ten, avoid, lead, false, cially, gions, duce, checker, board, stead, col, ors, thereby, just, mean, signed, fuzzy, placed, plat, lcd, tates, small, dis, crete, along, fined, tra, tory, trolled, ser, vos, nected, setup, against, process, moved, again, all, different, checkerboard, instead, ones, eliminate, reflection, prac, ema, dus, try, matog, pher, blue, vfx, pert, places, soft, ware, uses, sig, nif, cant, semi, hair, blur, vided, grounds, ables, trary, nev, tech, nique, guar, tee, cause, sumes, sent, very, pear, foundry, hol, wood, image, region, result, significantly, preserving, details, thresh, old, gen, ated, nar, row, lated, which, vides, hand, made, cut, ergy, miza, sists, plate, seven, scribed, posed, var, ous, bal, lished, hid, rest, ness, first, lic, tains, scat, rat, bles, avail, player, equipped, zoom, gion, lieve, help, ist, aid, official, opening, after, effects, online, full, mode, codes, matte, general, ranking, table, multiple, drop, down, menu, top, left, corner, switch, description, hairs, metrics, temporal, coherency, chosen, careful, generation, looking, accurate, owing, correction, lightning, during, capturing, improvement, website, loading, speed, interface, estimation, completion, videocompletion, objective,

Text of the page (random words):
oth im age and video mat ting meth ods we ap ply each mat ting method to the videos in our data set and then com pare the re sults us ing the fol low ing met rics of per pixel ac cu racy and tem po ral co herency look into our pa per 3 for com par i son of dif fer ent met rics here de notes to tal num ber of pix els and de note trans parency val ues of video mat ting un der con sid er a tion and ground truth cor re spond ingly at pixel of frame and de notes mo tion vec tor at pixel we use op ti cal flow al go rithm 11 com puted for ground truth se quences it is worth not ing that mo tion aware met rics will not give un fair ad van tage to mat ting meth ods based on the sim i lar mo tion es ti ma tion method since they do not have ground truth se quence the de tailed de scrip tion of used qual ity met rics is given in 3 public sequences for the train ing pur poses we pub lish here three test se quences with their ground truth trans parency maps de vel op ers and re searchers are wel come to use these se quences but we ask to cite us 3 castle alex dmitriy participate we in vite de vel op ers of video mat ting meth ods to use our bench mark we will eval u ate the sub mit ted data and re port scores to the de vel oper in cases where the de vel oper specif i cally grants per mis sion we will pub lish the re sults on our site we can also pub lish anony mous scores for blind re viewed pa pers to par tic i pate sim ply fol low these steps download the data set containing our sequences city flowers concert rain snow vitaliy artem slava juneau woods apply your method to each of our test cases upload the alpha and foreground sequences to any file sharing service we kindly ask you to maintain these naming and directory structure conventions if your method doesn t explicitly produce the foreground images you can skip uploading them in this case we will generate them using method proposed in 7 fill in this form to provide information about your method con tact us by email with any ques tions or sug ges tions at ques tions video mat ting com cite us to re fer to our eval u a tion or test se quences in your work cite our pa per 3 inproceedings erofeev2015 title perceptually motivated benchmark for video matting author mikhail erofeev and yury gitman and dmitriy vatolin and alexey fedorov and jue wang year 2015 month september pages 99 1 99 12 articleno 99 numpages 12 booktitle proceedings of the british machine vision conference bmvc publisher bmva press doi 10 5244 c 29 99 isbn 1 901725 53 7 url https dx doi org 10 5244 c 29 99 evaluation rating trimap size narrow medium wide trimap available for each frame year rank city rain concert flowers snow slava vitaliy artem juneau woods bayesian matting 2 2001 12 6 58 76 14 51 66 12 10 79 8 144 53 13 30 36 13 53 28 13 43 70 13 70 50 14 96 69 12 140 12 14 robust matting 12 2007 8 2 37 54 10 26 15 4 9 61 4 88 15 7 18 52 8 33 31 10 34 15 7 46 91 11 78 56 11 80 41 10 refine edge 15 2013 9 9 37 78 11 38 48 10 8 64 2 129 56 11 27 47 12 31 36 9 36 65 11 38 75 8 102 02 13 94 50 12 closed form 7 2008 7 9 31 34 6 31 80 7 10 96 10 149 48 14 15 86 5 23 67 7 30 15 5 30 60 5 69 94 9 82 13 11 learning based 13 2009 5 9 29 98 4 33 08 8 10 84 9 134 82 12 15 23 4 23 35 5 31 46 6 30 20 3 58 38 4 56 60 4 nonlocal matting s 5 2011 14 7 54 21 13 92 37 15 33 86 15 243 47 16 45 13 15 64 05 15 61 75 15 67 30 13 135 08 15 171 79 15 shared matting 4 2010 8 9 36 79 9 46 72 11 10 30 7 87 64 6 24 41 10 35 18 11 35 26 9 45 23 10 68 52 8 67 17 8 comprehensive sampling s 10 2008 7 9 34 48 8 30 18 6 12 10 11 79 60 4 16 17 6 25 02 8 35 25 8 41 05 9 77 72 10 70 18 9 knn matting 1 2012 13 9 67 83 15 92 22 14 64 73 16 112 46 10 32 96 14 61 50 14 50 87 14 79 20 15 113 41 14 123 63 13 spectral matting s 6 2012 15 5 86 32 16 105 24 16 13 15 12 210 04 15 48 76 16 65 72 16 68 83 16 109 44 16 157 18 16 182 82 16 sparse sampling s 14 2016 6 8 33 42 7 33 50 9 9 94 5 81 06 5 18 42 7 23 64 6 35 83 10 38 00 7 65 17 6 64 12 6 deep matting 16 2016 4 6 27 33 3 21 40 3 20 33 13 44 93 2 19 23 9 19 99 3 30 00 4 30 55 4 46 18 3 46 39 2 self adaptive 17 2016 4 8 30 05 5 29 30 5 8 02 1 96 44 9 14 32 3 21 96 4 29 92 3 35 97 6 66 05 7 61 49 5 information flow 20 2017 10 6 50 81 12 59 62 13 24 80 14 94 81 8 25 93 11 48 82 12 42 14 12 55 32 12 64 02 5 66 56 7 background matting 21 2020 2 5 22 77 2 17 53 2 10 06 6 63 61 3 10 80 2 17 15 2 25 19 1 22 27 2 45 44 2 46 85 3 fba matting 22 2020 1 3 17 25 1 16 69 1 8 76 3 40 94 1 9 92 1 15 30 1 27 24 2 19 10 1 28 67 1 30 45 1 bayesian matting 2 2001 13 6 73 18 14 67 92 13 16 06 11 157 28 14 39 59 14 60 37 13 56 05 14 85 53 15 115 88 14 169 33 14 robust matting 12 2007 9 5 48 19 11 32 89 7 14 72 9 99 81 7 25 96 9 42 51 10 38 88 9 59 08 11 93 79 11 96 13 11 refine edge 15 2013 9 6 42 23 10 41 74 10 10 45 2 127 02 11 31 40 12 35 08 9 39 46 10 42 91 8 108 71 12 97 76 12 closed form 7 2008 7 6 34 53 6 35 86 8 13 35 6 154 12 13 21 64 8 27 19 7 33 01 5 35 45 5 81 48 8 92 75 10 learning based 13 2009 6 2 33 44 5 37 63 9 13 23 5 138 31 12 19 95 7 26 51 5 34 51 6 34 70 4 68 41 5 63 53 4 nonlocal matting s 5 2011 14 5 55 04 13 97 85 14 31 81 14 242 75 16 46 94 15 67 11 15 65 09 15 69 01 13 144 66 15 176 61 15 shared matting 4 2010 9 3 41 44 9 56 75 11 14 14 7 96 71 6 27 66 11 42 92 11 40 05 11 58 17 10 83 02 9 78 91 8 comprehensive sampling s 10 2008 6 9 38 43 8 28 29 4 14 45 8 83 85 4 16 84 4 27 29 8 37 07 7 47 38 9 89 96 10 78 45 7 knn matting 1 2012 13 5 79 35 15 98 71 15 66 24 16 112 81 10 34 60 13 64 17 14 54 44 12 84 38 14 112 13 13 128 86 13 spectral matting s 6 2012 15 3 99 05 16 125 29 16 16 04 10 214 42 15 59 61 16 74 55 16 77 25 16 122 89 16 162 73 16 188 16 16 sparse sampling s 14 2016 5 2 33 20 4 31 92 5 12 34 4 85 52 5 17 66 5 24 64 4 37 78 8 39 87 6 71 05 6 69 94 5 deep matting 16 2016 4 1 28 67 3 21 10 3 41 80 15 44 81 2 16 57 3 21 04 3 30 57 3 30 94 3 46 40 3 50 27 3 self adaptive 17 2016 6 1 34 71 7 32 21 6 11 02 3 101 76 9 18 61 6 26 53 6 31 23 4 40 01 7 75 29 7 70 68 6 information flow 20 2017 10 5 54 68 12 67 36 12 27 93 13 100 44 8 26 93 10 55 24 12 55 54 13 60 15 12 66 35 4 79 33 9 background matting 21 2020 3 0 22 97 2 18 15 2 16 25 12 68 75 3 10 85 2 17 08 2 24 61 1 22 74 2 43 41 2 46 86 2 fba matting 22 2020 1 1 18 90 1 16 87 1 9 85 1 42 48 1 10 16 1 15 64 1 27 35 2 19 85 1 28 80 1 30 95 1 bayesian matting 2 2001 14 1 98 13 15 95 95 13 38 54 14 170 76 14 59 51 15 80 72 15 67 64 13 122 08 14 135 59 14 179 25 14 robust matting 12 2007 10 7 67 71 13 49 02 9 19 87 7 109 79 9 44 10 13 62 18 11 50 54 10 81 49 12 118 09 12 115 96 11 refine edge 15 2013 9 1 50 31 7 51 81 10 15 85 4 125 25 11 40 73 12 45 38 9 46 43 9 52 22 7 118 88 13 103 30 9 closed form 7 2008 7 8 45 04 6 41 69 6 24 19 9 159 47 13 30 09 9 36 85 8 39 20 4 46 97 5 104 07 8 110 26 10 learning based 13 2009 6 4 44 97 5 43 75 7 21 90 8 143 54 12 26 50 6 33 89 6 40 44 6 45 93 4 90 35 6 81 22 4 nonlocal matting s 5 2011 13 7 56 76 10 113 90 15 33 97 13 241 31 16 50 74 14 77 14 14 68 37 14 76 43 11 159 08 15 181 14 15 shared matting 4 2010 9 7 55 84 9 79 09 11 19 78 6 107 45 8 32 94 10 60 78 10 51 47 11 122 33 15 107 34 9 97 22 8 comprehensive sampling s 10 2008 6 9 56 79 11 37 36 5 16 96 5 87 64 4 24 04 5 34 32 7 43 69 7 64 63 9 110 50 10 93 35 6 knn matting 1 2012 12 7 95 89 14 112 70 14 66 12 16 113 10 10 37 18 11 71 26 13 60 03 12 98 42 13 110 87 11 138 66 13 spectral matting s 6 2012 15 3 116 87 16 157 37 16 26 12 10 218 94 15 76 22 16 88 74 16 91 69 16 148 70 16 169 33 16 193 46 16 sparse sampling s 14 2016 4 8 43 32 4 34 66 4 14 78 3 89 79 5 19 23 4 29 92 4 44 59 8 51 94 6 85 44 5 81 77 5 deep matting 16 2016 4 0 31 29 3 23 16 2 53 02 15 44 93 2 18 20 3 24 36 3 32 04 3 37 60 3 50 48 3 54 76 3 self adaptive 17 2016 6 4 51 29 8 46 56 8 14 35 2 106 71 7 27 55 7 31 98 5 39 35 5 54 80 8 95 90 7 95 73 7 information flow 20 2017 10 3 66 86 12 86 01 12 30 99 12 104 14 6 29 25 8 64 32 12 82 92 15 69 33 10 69 71 4 117 38 12 background matting 21 2020 3 0 25 77 2 23 62 3 29 33 11 75 90 3 11 74 2 18 31 2 24 83 1 24 97 2 43 92 2 50 51 2 fba matting 22 2020 1 1 22 02 1 19 82 1 11 94 1 44 62 1 11 21 1 17 05 1 28 11 2 22 14 1 29 79 1 33 06 1 bayesian matting 2 2001 13 8 35 93 14 65 24 12 9 28 10 104 68 15 27 61 13 35 04 16 46 41 13 60 34 15 88 46 15 121 91 15 robust matting 12 2007 8 2 22 32 9 32 73 4 7 26 4 47 27 13 16 04 8 18 72 10 34 47 7 24 96 8 53 42 11 42 40 8 refine edge 15 2013 7 2 14 41 2 44 08 9 6 30 3 40 69 8 18 36 10 16 53 6 34 93 8 17 83 4 53 43 12 43 60 10 closed form 7 2008 6 2 16 37 4 41 29 7 7 67 6 35 14 7 14 59 5 16 40 5 32 43 5 17 37 3 51 40 8 47 71 12 learning based 13 2009 5 8 17 29 6 42 99 8 7 88 8 34 03 5 14 44 4 17 07 8 34 37 6 18 37 5 46 02 4 39 46 4 nonlocal matting s 5 2011 14 6 29 07 13 110 76 14 14 10 15 120 01 16 29 82 15 32 69 14 59 91 15 36 22 14 108 43 16 96 82 14 shared matting 4 2010 9 2 22 72 10 59 59 11 7 72 7 45 92 12 20 34 11 22 35 11 38 02 11 29 56 11 47 45 5 38 19 3 comprehensive sampling s 10 2008 8 9 22 78 11 39 97 6 13 83 14 33 11 4 15 61 6 17 51 9 37 09 9 25 72 9 53 33 10 45 04 11 knn matting 1 2012 13 7 36 16 15 115 08 15 25 58 16 44 09 10 27 80 14 31 61 13 56 08 14 34 80 13 87 53 14 75 36 13 spectral matting s 6 2012 13 9 65 50 16 131 74 16 4 78 1 103 61 14 39 03 16 34 39 15 63 21 16 64 40 16 87 44 13 122 53 16 sparse sampling s 14 2016 8 9 21 78 8 44 85 10 9 86 11 34 07 6 16 94 9 16 55 7 37 40 10 25 84 10 51 77 9 42 98 9 deep matting 16 2016 4 7 16 79 5 25 93 3 13 12 13 26 74 2 15 64 7 14 58 3 31 85 3 19 25 6 45 03 3 37 27 2 self adaptive 17 2016 5 8 18 76 7 38 89 5 7 37 5 45 75 11 13 36 3 15 44 4 32 26 4 20 12 7 49 15 6 40 48 6 information flow 20 2017 10 8 28 00 12 78 55 13 10 64 12 41 13 9 22 56 12 28 71 12 42 89 12 29 75 12 49 89 7 41 44 7 background matting 21 2020 2 4 15 00 3 21 44 2 6 25 2 32 74 3 10 53 2 13 40 2 27 79 1 16 29 2 42 39 2 40 10 5 fba matting 22 2020 1 9 12 65 1 19 55 1 8 79 9 23 72 1 10 30 1 12 32 1 28 52 2 14 82 1 29 27 1 28 06 1 bayesian matting 2 2001 14 2 46 96 15 77 27 12 11 32 10 111 60 15 33 85 15 39 54 16 56 86 12 75 50 16 102 70 15 151 60 16 robust matting 12 2007 9 3 26 36 10 37 73 5 8 35 6 52 07 13 19 18 9 21 45 10 37 08 8 28 55 10 57 78 12 47 95 10 refine edge 15 2013 6 4 15 24 3 47 35 10 6 58 2 40 54 8 19 40 10 16 77 5 35 74 6 18 24 4 54 11 10 43 43 6 closed form 7 2008 6 8 17 25 5 42 80 8 7 11 4 35 79 7 16 70 8 17 10 7 33 94 5 18 08 3 52 82 9 48 94 12 learning based 13 2009 5 9 18 97 6 45 14 9 7 45 5 34 62 3 16 32 7 18 09 8 36 77 7 19 50 6 48 00 4 39 16 4 nonlocal matting s 5 2011 14 1 29 67 12 114 94 14 14 84 13 121 99 16 31 03 14 33 66 14 61 34 15 37 04 13 117 02 16 93 18 14 shared matting 4 2010 9 8 24 21 9 68 25 11 9 00 8 49 95 12 21 88 11 24 25 11 42 74 11 37 14 14 51 50 6 39 51 5 comprehensive sampling s 10 2008 8 8 26 80 11 35 36 4 15 59 14 34 80 5 15 73 4 18 29 9 39 17 10 27 81 9 55 93 11 48 73 11 knn matting 1 2012 13 2 38 85 14 120 08 15 26 54 16 44 01 9 28 25 13 31 83 13 57 58 13 35 78 12 85 37 14 75 34 13 spectral matting s 6 2012 13 7 72 45 16 155 21 16 5 03 1 104 95 14 44 50 16 35 84 15 64 96 16 66 66 15 85 23 13 130 32 15 sparse sampling s 14 2016 7 2 23 66 8 39 99 6 11 54 11 35 17 6 16 17 6 16 48 4 38 93 9 27 46 8 51 93 7 44 55 7 deep matting 16 2016 4 3 17 21 4 24 74 3 21 03 15 27 68 2 14 33 3 14 98 3 31 41 3 19 05 5 45 10 3 36 96 2 self adaptive 17 2016 7 0 20 61 7 40 20 7 8 93 7 46 73 11 15 80 5 17 03 6 32 95 4 21 69 7 52 15 8 44 90 8 information flow 20 2017 11 1 34 10 13 86 71 13 13 95 12 45 99 10 23 26 12 31 00 12 61 24 14 30 42 11 49 38 5 44 99 9 background matting 21 2020 2 3 14 92 2 22 50 2 6 91 3 34 67 4 10 69 2 13 21 2 27 25 1 16 16 2 40 04 2 38 86 3 fba matting 22 2020 1 9 13 67 1 19 98 1 9 62 9 24 32 1 10 47 1 12 41 1 28 71 2 15 11 1 29 15 1 28 29 1 bayesian matting 2 2001 15 2 68 32 15 110 78 13 29 70 16 119 54 15 47 66 15 52 24 16 67 60 15 118 01 16 98 34 15 148 17 16 robust matting 12 2007 9 8 32 72 11 53 12 8 9 85 5 56 83 13 25 40 12 27 41 10 42 73 8 36 49 11 62 55 12 52 27 8 refine edge 15 2013 5 8 15 82 3 57 57 10 6 58 2 40 25 8 20 67 9 18 04 4 37 30 6 19 11 3 55 24 7 44 18 6 closed form 7 2008 5 9 18 75 5 48 30 6 7 76 3 36 78 6 19 57 7 18 21 6 35 50 4 19 15 4 57 13 9 53 95 9 learning based 13 2009 5 6 21 35 6 50 91 7 7 84 4 35 59 3 18 92 5 19 37 8 39 75 7 21 31 6 51 66 5 41 99 5 nonlocal matting s 5 2011 13 7 31 25 10 129 13 14 17 96 12 123 49 16 31 74 14 36 13 15 62 32 13 38 82 13 127 06 16 91 37 14 shared matting 4 2010 10 2 27 95 9 93 58 11 10 60 9 55 32 12 25 04 11 27 72 11 48 28 11 65 06 14 58 63 10 41 34 4 comprehensive sampling s 10 2008 8 8 34 62 12 42 66 5 16 53 11 35 69 4 19 23 6 21 26 9 44 33 10 33 28 10 60 01 11 55 61 10 knn matting 1 2012 12 8 42 71 14 130 14 15 26 50 15 43 95 9 29 19 13 33 06 12 59 22 12 38 04 12 82 47 13 76 16 13 spectral matting s 6 2012 13 5 81 14 16 183 76 16 5 49 1 108 67 14 54 31 16 35 81 14 67 31 14 70 45 15 82 74 14 128 41 15 sparse sampling s 14 2016 6 8 27 39 8 41 55 4 12 15 10 36 95 7 17 04 4 18 19 5 43 94 9 32 67 8 54 81 6 49 26 7 deep matting 16 2016 4 1 18 72 4 28 24 2 23 65 14 28 18 2 15 36 3 16 66 3 31 80 3 20 98 5 47 83 3 36 95 2 self adaptive 17 2016 7 9 25 66 7 56 94 9 10 23 6 49 29 10 19 97 8 18 71 7 37 00 5 25 31 7 55 89 8 57 11 12 information flow 20 2017 11 2 41 09 13 106 61 12 18 03 13 49 30 11 24 59 10 35 49 13 94 92 16 32 86 9 49 22 4 56 92 11 background matting 21 2020 2 9 15 57 2 28 72 3 10 58 8 36 50 5 11 33 1 13 56 2 27 23 1 16 71 2 39 75 2 38 14 3 fba matting 22 2020 1 8 14 43 1 23 68 1 10 55 7 24 98 1 11 49 2 12 93 1 29 29 2 15 85 1 29 64 1 29 22 1 bayesian matting 2 2001 13 8 1 85 14 3 58 12 0 12 11 13 33 14 0 74 14 2 31 16 1 47 13 4 51 15 8 01 14 16 23 15 robust matting 12 2007 9 5 0 89 11 0 94 4 0 07 4 4 82 13 0 25 9 0 88 10 0 77 8 1 58 12 3 89 12 3 92 12 refine edge 15 2013 8 0 0 55 6 1 48 10 0 06 2 3 99 10 0 34 10 0 54 9 0 70 7 0 80 6 3 80 11 3 16 9 closed form 7 2008 5 7 0 49 3 1 35 8 0 09 8 2 88 4 0 19 5 0 35 6 0 61 3 0 61 4 2 75 6 3 38 10 learning based 13 2009 5 6 0 50 5 1 42 9 0 09 9 2 88 5 0 18 4 0 34 5 0 68 6 0 64 5 2 37 4 2 43 4 nonlocal matting s 5 2011 14 9 1 56 13 8 17 15 0 70 15 20 97 16 1 11 15 1 88 15 2 51 16 2 35 14 11 17 16 13 37 14 shared matting 4 2010 10 1 0 79 10 2 73 11 0 09 7 4 50 12 0 35 11 1 00 12 0 91 11 1 43 11 2 94 8 2 94 8 comprehensive sampling s 10 2008 8 8 0 74 9 1 06 5 0 19 12 3 15 7 0 22 7 0 39 8 0 86 10 1 24 9 3 65 10 3 40 11 knn matting 1 2012 13 2 2 28 15 6 40 14 1 22 16 3 47 8 0 61 13 1 02 13 1 73 14 1 94 13 6 42 13 7 73 13 spectral matting s 6 2012 14 5 7 03 16 13 25 16 0 08 6 15 36 15 2 09 16 1 84 14 2 45 15 6 59 16 8 71 15 19 97 16 sparse sampling s 14 2016 7 9 0 66 8 1 30 7 0 11 10 3 09 6 0 25 8 0 38 7 0 86 9 1 04 8 2 94 9 2 76 7 deep matting 16 2016 4 6 0 50 4 0 48 3 0 32 14 1 15 2 0 22 6 0 25 3 0 63 5 0 57 3 1 84 3 1 84 3 self adaptive 17 2016 5 6 0 58 7 1 15 6 0 06 1 4 43 11 0 16 3 0 33 4 0 62 4 0 87 7 2 91 7 2 66 6 information flow 20 2017 10 2 1 10 12 3 94 13 0 27 13 3 90 9 0 41 12 0 97 11 1 10 12 1 38 10 2 46 5 2...