［ソフトウェア］Saliency Map Algorithm：MATLAB Source Code

http://www.cnblogs.com/ToDoToTry/archive/2011/08/09/2133112.html
［ソフトウェア］Saliency Map Algorithm：MATLAB Source Code
Below is MATLAB code which coputtes a salience/saliency map for an image or image sequence/video(eigther Graphh-Based Visual Saliency(GBVS) or the standar Itti，Kuch，Niebur PAMI 1998 saliency map）.See the includ readme file for details.I also have a newer，simpler version implement ing only the Itti algorithm(see) simpsal/readme.txt).
Please try it out and email 私 if you have any questions.
Download:Please select one of the follwing to download:
[gbvs.zip]This package includes an implementation of the full GBVS algorithm.It also lets you compute an Itti，Kuch，Niebur saliency map.It includes a function for computting the ROC score between eye-movements(provided as x，y fixation points)and a saliency map，and a function for displaying a saliency map overlayed on top of an image(as in the screenshow)  Right after you download the zip file，you must change into the gbvs/directory and run：>> gbvs_install  You only need to run that the first time.Afterwards，you can generate a saliency map as follows:To load an mage:>> img = imread('samplepics/1.jpg');
To compute a GBVS map:>> map = gbvs(img); % map.master_map contains the actual saliency map  To compute an Itti-Niebur map:>> map_itti = ittikochmap(img); % map_itti.master_map contains the actual saliency map  And you can visuize the saliencmap tomap
[simpsal.zip]This is a radically simplified version of the code implement ing only the Itti algorithm.See シンプル/readme.txt for details.This is is useful if you want to edit something minmail for reearch purposes.More feature re re re re included in the GBVS package above.  Note 1：if you want the out put saliency map to be the same size as the inputイメージ、cal simpsal like this：>> show_imgnmap( img , map );  Note 2： If you are looking for a very quick&dirty saliency map、call simpsal with default_fast_param(e.g.simpsal)fast_param). The output of simpsal is not identical to that put by the Itti algorithm option from the GBVS package.simpsal is a bit faster and accurate(in fixation predictions)than the Itti Ittimplement able.Superover.Surrine.Surisethe first one is a more faithful reproduction of the ori ginal algorithm(exact mechaism of center surround surround subtraction and scale parameters and selection).However，they aridentical in spirit and structure and structure and vantractity  Note:If you install GBVS、make sure you put it in a directory the full path to which does not contain spaces.Note:If you have trouble/run into an error running imshow，imresize，prctile，or mat 2 gray，try downloading studtools.zip and put the contensts in the directory containing the saliency code.Last udated Markch 15,2011.Copyright：The Itti、Kuch、Niebur algorithm is provided for education al use only and for copyright information on that please read this.The GBVS algorithm can be used for any purposes、but only if it is fully cited/acknowledged. Reference:

See the 

oricginnal paper

 by 

ジョナサンHarel

.A 

poster

 is also available.

Please cite use of this code as:
-J.Harel，C.Koch，and P.Perona，「Graphh-Based Visual Saliency」，Proceding s of Neural Information Processing Systems（NIPS），2006. [pdf
-and/or-
-J.Harel，A Saliency Implementation in MATLAB：http://www.klab.caltech.edu/~hall/share/gbvs.php
Feel free to contact 私 via email with any questions. シュレンショット:  My Interpretation of why GBVS is more preditive than Itti:Two very trivial factors which greatly affect the measred predition accuracy of saliency algorithms(up 30%improvement in some caser)are e e e therever fireceand(2)extentt of center bias(for instance，a global centr biasmultilicativelyaaplind to the matermap).So，inhent variaaaaabilityininininininininstststt t t t t to the repopoted the repopoted performancinininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininininon a database of 200 face mages you will find 0.841 mean ROC for GBVS and 0.812 mean ROC for Itti(NSS values 1.321 and 1.228 rep.)after optimizing for blur andセンターBias.Based on n n n n n inspection of may maps、I believe that this remaning performance gain is due to long-range copetition in the GBVS algorithm which is not captured accurately enough by the Itti algorithm.  -JH。