Matching signs Using SURF

I am trying to match signs of American Sign Language using SURF.

Ideally, it should match with "C" in database images but it is not so.

I am attaching code.

Currently, input image is CC.jpg

I have to make such a code that should work different input images like AA.jpg, BB.jpg, CC.jpg, DD.jpg and so on.

INPUT IMAGES: http://www.2shared.com/file/1X6p62-T/input_images.html

Answer 1

[Bruce Tannenbaum]

Hi,

This is similar to the example Detecting Objects in a Cluttered Scene With Point Feature Matching.

First, make sure you are detecting enough features in both CC.jpg and database_labelled.jpg. You can detect more features using the "MetricThreshold" argument on detectSURFFeatures. Then, test different levels of the "MatchThreshold" argument on matchFeatures. Sometimes the default settings are not sufficient.

Also, is there enough resolution to database_labelled.jpg to detect at least 20+ features for each sign configuration? If not, this will significantly impair your ability to use features to recognize the sign.

Hope this helps,

Bruce Tannenbaum

MathWorks

[Explorer]

I have just tried it by including MetricThreshol but still there is same problem.

[Explorer]

I want to do something like thing like the one as shown in video. http://vimeo.com/55812802

[Walter Roberson]

You did notice that the description of the video says "Two semester's work" ?

[Explorer]

Okay, what will you suggest me to resolve my problem?

[Walter Roberson]

Patient: "Doctor, Doctor, it hurts when I do (this)!"
Doctor: "Well then, don't do that."

[Explorer]

hahahaha...