Which image comparison method should I use?

[ilkery]

Hi All,

I have the same images (the same pattern) those are taken from different TV panels via LVDS. (This is not an identical process, let's say we have 2 images.)

The patterns (color bar) are the same.But it fails when I compare them.

How can I compare them?

Which filters or methodology should be used.

Could you please inform meon these subjects.

Thanks in advance. 

[Erik-S]

Hello,

It looks like the method you are using will compare each individual pixel. Since you are taking two different images, it will be nearly impossible to get an exact match on a pixel by pixel level. Instead, I would try using a pattern match or edge detection. This will give you some room for error so that you can pass certain image pairs if they match within a certain threshold. Here is a link to some of the available algorithms in the Vision Development Module. You may be interested in the Golden Template Comparison, which compares pixels, but allows some specified wiggle room.

You may also want to post this question on the Machine Vision discussion forums since this is a Vision specific application. You will have a better chance of someone who has made a similar application viewing your question.

-Erik S