Fuzzy matching parts of the image

I need to determine the existence of a smaller image inside a larger image.

The match should be fuzzy, and it is desirable that I know how much it matches (%), but I can calculate the accuracy of myself after the match is completed, if necessary.

My goal is to compare the FFT (Fast Fourier Transfrom) spectrogram (frequency visualization in music) with the original music track. The small image that I compare is just a subset (both in time and in the frequency range) of the original track (for example, the cutout of the image below).

Where to begin? Are the same algorithms used for object recognition for this task?

I am primarily looking for C # /. NET libraries / samples, but also information on implementations and problems / traps.

I am considering using artificial neural networks for recognition training. Any thoughts?

An example of which images I want to match might look like this: http://img812.imageshack.us/img812/548/musicspectogram2.png

+3
source share
2 answers

I think that to consider this problem as a problem of pattern recognition is to ignore the basic structure of the problem. In particular, you can see how Shazam fixes the problem. There are some interesting links to Quora in this question:

http://www.quora.com/How-does-Shazam-work

-, , . , , , . .

-, Slate, , , , .

0

. .

Mathematica:

enter image description here

, , .

: Wavelet, .

0

All Articles