how to scan image for specific pattern

i want to scan a binary image row by row for a consistent pattern of black:white:black:white:black to the scale of 1:1:3:1:1 pixels respectively. i want to use it for QR code scanning. something like this

Answer 1

[Anand]

Use normalized cross correlation (NCC) provided by the normxcorr2 function.

You can try something like this:

%the pattern you're looking for is B W B B B W B
template = logical([0 1 0 0 0 1 0]);

%find NCC of template with image BW
cc = normxcorr2(template,BW);

Then find peaks in cc to locate co-ordinates of the pattern.

[Image Analyst]

I thought of that first also but then I realized that it wouldn't work if the QR bar code in the image isn't the same size and pattern as the pattern he's searching for. I believe he wants to locate any QR code, not just some code of a pattern and size that is known in advance. So if he can't specify in advance what to look for then he'll have to look to texture segmentation. Or else search the literature for papers like I mentioned in my comment.

[Anand]

Yes, I agree. Bur since genesis mentioned that the scale specifically as 1:1:3:1:1 "pixels" I thought he might know more about his image data.

Answer 2

[Image Analyst]

Try using stdfilt() to look for areas of high texture in the image.

[Image Analyst]

You might try this:

Kato, Y.[Yuji], Deguchi, D.[Daisuke], Takahashi, T.[Tomokazu], Ide, I.[Ichiro], Murase, H.[Hiroshi], Low Resolution QR-Code Recognition by Applying Super-Resolution Using the Property of QR-Codes, ICDAR11(992-996).

By the way, if you do want to play around with normxcorr2 (which probably won't work in your situation) you can try the demo I attached below.