Pattern matching

[johnmerthyr32]

I want to use a normalised correlation function for pattern matching in 16 bit greyscal images. Is this possible to do without reducing the images to 8 bits?

[Klemen]

Hello,

I belive that NI Vision only supports U8 image types. You would need write your own code to perform this.

Is there any particular reason that you want to perform the matching on the 16-bit image?

Best regards,

K

[johnmerthyr32]

Thanks for your comments.

I am looking at sets of images that have different exposure conditions and much of the important information is in pixels of lowest intensity. Converting images to 8 bits would considerably reduce the dynamic range and normalised correlation for template tracking would have more errors.

Best regards

J

[Klemen]

Hello,

I understand...

So you would need to code this yourself or use the OpenCV template matching. The function can take 32-bit floating point image in addition to 8-bit gray.

You can check the example on how to call the .dll for template matching in Labview. The code is for 8-bit images,but you can easily modify for 32-bits. So if you convert your 16- bit to 32-bit you would not lose the dynamic range.

Check the link in my signature below for the example.

Best regards,
K