extract features from image

[rezanasiripour]

Hi
I should implementation system of object retrieval and now need to extract feature.
have you plans of extract feature?for example SIFT,HOG,surf or .....
If you're mail me.
Thanks

rezanasiripour@yahoo.com

 

[tgootee]

Feature extraction is a very broad subject area!

Depending on your application, it might be good to focus on using scale-invariant, rotation-invariant, and illumination-independent features.

Do you have a target hardware platform, software development environment, and programming language?

Many of the relevant scientific papers are only available on pay-to-subscribe sites, such as IEEE Explore. But there are a lot of free PDFS and downloadable code, and free software toolboxes at the links below.

Here are some Pattern Recognition and Machine Vision links that I have collected, recently:

(The link below has some Mathcad code (fairly elementary but extremely useful for beginners). Fully-functional non-expiring Mathcad 15 can be obtained by buying "Essential Mathcad", 2nd edition, by Brent Maxfield, at amazon.com. It comes with a license key that enables a 259 MB download of the software. It's a nice way to get a $1500 software package for about $40. The book is, I believe, out of print, now. But there were still new copies available on Amazon, the last time I checked, a few months ago.)
http://users.ecs.soton.ac.uk/msn//book/

OpenCV is a free multi-platform software library for computer vision:

**broken link removed**
http://docs.opencv.org/modules/features2d/doc/features2d.html
http://docs.opencv.org/doc/tutorial...udio_Opencv.html#windows-visual-studio-how-to
http://www.embedded-vision.com/intr...using-opencv?gclid=COKD8-Sc7rYCFeU-Mgodll8AFw
http://stackoverflow.com/questions/...-and-tracking-based-on-feature-detection?rq=1
http://answers.opencv.org/questions/scope:all/sort:activity-desc/page:1/
http://sourceforge.net/projects/opencvlibrary/files/?source=navbar
http://stackoverflow.com/questions/...tion-performance-improvement-with-opencv?rq=1

More-general types of Machine Vision and Pattern Recognition links:

http://www.visionbib.com/bibliography/contents.html
http://www.cs.cmu.edu/afs/cs/project/cil/ftp/html/vision.html
**broken link removed**
**broken link removed**
http://www4.comp.polyu.edu.hk/~cslzhang/code.htm
http://orfeo-toolbox.org/SoftwareGuide/SoftwareGuidepa3.html#SoftwareGuidech14.html
**broken link removed**
**broken link removed**
**broken link removed**
**broken link removed**
http://www.mathworks.com/matlabcentral/fileexchange/authors/31581
**broken link removed**
http://www.vision.jhu.edu/code/
http://www.mathworks.com/matlabcentral/fileexchange/37933-sfta-texture-extractor
http://www.robots.ox.ac.uk/~vgg/research/affine/detectors.html
http://www.cs.cmu.edu/~cil/v-pubs.html
http://www.cs.cmu.edu/~cil/v-source.html
http://homepages.inf.ed.ac.uk/rbf/CVonline/#nonwiki
**broken link removed**
http://videolectures.net/cvpr2010_deselaers_gess/
http://homepages.inf.ed.ac.uk/rbf/HIPR2/
http://homepages.inf.ed.ac.uk/rbf/IAPR/researchers/index.php
http://www.macs.hw.ac.uk/texturelab/files/publications/
http://www.macs.hw.ac.uk/texturelab/resources/links/
http://www.cs.hmc.edu/~fleck/computer-vision-handbook/index.html
http://www.cs.cmu.edu/~cil/vision.html
http://homepages.inf.ed.ac.uk/rbf/IAPR/researchers/D2PAGES/d2tut.htm
**broken link removed**

Cheers,

Tom

 

[rezanasiripour]

Feature extraction is a very broad subject area!

Depending on your application, it might be good to focus on using scale-invariant, rotation-invariant, and illumination-independent features.

Do you have a target hardware platform, software development environment, and programming language?

Many of the relevant scientific papers are only available on pay-to-subscribe sites, such as IEEE Explore. But there are a lot of free PDFS and downloadable code, and free software toolboxes at the links below.

Here are some Pattern Recognition and Machine Vision links that I have collected, recently:

(The link below has some Mathcad code (fairly elementary but extremely useful for beginners). Fully-functional non-expiring Mathcad 15 can be obtained by buying "Essential Mathcad", 2nd edition, by Brent Maxfield, at amazon.com. It comes with a license key that enables a 259 MB download of the software. It's a nice way to get a $1500 software package for about $40. The book is, I believe, out of print, now. But there were still new copies available on Amazon, the last time I checked, a few months ago.)
http://users.ecs.soton.ac.uk/msn//book/

OpenCV is a free multi-platform software library for computer vision:

**broken link removed**
http://docs.opencv.org/modules/features2d/doc/features2d.html
http://docs.opencv.org/doc/tutorial...udio_Opencv.html#windows-visual-studio-how-to
http://www.embedded-vision.com/intr...using-opencv?gclid=COKD8-Sc7rYCFeU-Mgodll8AFw
http://stackoverflow.com/questions/...-and-tracking-based-on-feature-detection?rq=1
http://answers.opencv.org/questions/scope:all/sort:activity-desc/page:1/
http://sourceforge.net/projects/opencvlibrary/files/?source=navbar
http://stackoverflow.com/questions/...tion-performance-improvement-with-opencv?rq=1

More-general types of Machine Vision and Pattern Recognition links:

http://www.visionbib.com/bibliography/contents.html
http://www.cs.cmu.edu/afs/cs/project/cil/ftp/html/vision.html
**broken link removed**
**broken link removed**
http://www4.comp.polyu.edu.hk/~cslzhang/code.htm
http://orfeo-toolbox.org/SoftwareGuide/SoftwareGuidepa3.html#SoftwareGuidech14.html
**broken link removed**
**broken link removed**
**broken link removed**
**broken link removed**
http://www.mathworks.com/matlabcentral/fileexchange/authors/31581
**broken link removed**
http://www.vision.jhu.edu/code/
http://www.mathworks.com/matlabcentral/fileexchange/37933-sfta-texture-extractor
http://www.robots.ox.ac.uk/~vgg/research/affine/detectors.html
http://www.cs.cmu.edu/~cil/v-pubs.html
http://www.cs.cmu.edu/~cil/v-source.html
http://homepages.inf.ed.ac.uk/rbf/CVonline/#nonwiki
**broken link removed**
http://videolectures.net/cvpr2010_deselaers_gess/
http://homepages.inf.ed.ac.uk/rbf/HIPR2/
http://homepages.inf.ed.ac.uk/rbf/IAPR/researchers/index.php
http://www.macs.hw.ac.uk/texturelab/files/publications/
http://www.macs.hw.ac.uk/texturelab/resources/links/
http://www.cs.hmc.edu/~fleck/computer-vision-handbook/index.html
http://www.cs.cmu.edu/~cil/vision.html
http://homepages.inf.ed.ac.uk/rbf/IAPR/researchers/D2PAGES/d2tut.htm
**broken link removed**

Cheers,

Tom

hi again.
I can extract feature of 72 images(64 of images is training and 8 test).vector features is <72*1 cell>.
but now i do operation training with SVM(svm matlab).
program wrote is :

lbl=zeros(72,1);
trdata = b(1:64,1);
trlbl = lbl(1:64,1);
tstdata = b(65:72,1);
tstlbl = lbl(65:72,1);
trlbl1 = -ones(size(trlbl));
trlbl1(double(trlbl) == 1) = 1;
svm1 = svmtrain(trdata,trlbl1);
Group=svmclassify(svm1,tstdata);

but error is:
TRAINING must be a numeric matrix.
please help me.thanks

 

[tgootee]

You will have some algorithm programmed to extract feature vector or matrix from one image. You will run that for each training image and store the resulting calculated feature vectors or matrices. Then you will use that data to create a classification algorithm. Then your main program will take any test image and will apply your feature extraction algorithm to it and then your classifier algorithm will decide if the current test image is what you are looking for, or not.

You get to find out what types of features and what type of classifier will work the best, for your images, and your application.

 

[rezanasiripour]

I attached mfile and images(object retrieval.rar).
please help me
thanks