From: Aui on
Hello All,

Please could you explain how can I find the image matching positions once I know matching values for all pixels in the image and how to get the subpixel accuracy.

I know I might be in the wrong place but here there are a lot of image processing/computer vision experts. I am now working on my midterm for image matching with subpixel accuracy.

I have already implemented the code for cross correlation matching (not in matlab). Let say, my template exists 4 places in an image to search for. I expect that I could find only 4 points with cross-correlation-coef = 1 or close. However, as you can imagine, there are not only these 4 points that show good matched results but the neighborhood of these 4 points also results very high value like there are 4 hills.

Question:
=========
1. How can I find these 4 positions which represent the top of the hill?
2. After I got this position, how can I find the subpixel accuracy?

Thank you very much.
From: ImageAnalyst on
"Aui " <littlebearproj...(a)gmail.com>:

Pick your favorite method from those listed here:

http://iris.usc.edu/Vision-Notes/bibliography/match-pl495.html#Image%20Registration%20with%20Subpixel%20Accuracy

Each author probably claims his or hers is the best method.