Revision history [back]

In computer vision, what are technique/algorithms/ways to measure the homogeneity between 2 images? By homogeneity I mean measure in terms of:

• Colour/intensity for single channel
• Edginess (is that the term), ie, how rough or smooth it is.

As a practical usecase: I have an image that I have divided into grid cells. I wish to select a specific cell and search its neighbourhood (8 cells around it) and identify which of those neighbouring cells are most similar (similar homogeneity) to the subject cell.

I am aware of simple techniques like; calculate the mean intensity between images and measure the difference. But are there other more advanced techniques? I know SIFT finds keypoint descriptors in an image so I guess I could compute SIFT descriptors for each image then calculate the number of matches between them. Could I use an integral image here, histogram backprojection?

In computer vision, what are technique/algorithms/ways to measure the homogeneity between 2 images? By homogeneity I mean measure in terms of:

• Colour/intensity for single channel
• Edginess (is that the term), ie, how rough or smooth it is.

As a practical usecase: I have an image that I have divided into grid cells. I wish to select a specific cell and search its neighbourhood (8 cells around it) and identify which of those neighbouring cells are most similar (similar homogeneity) to the subject cell.

I am aware of simple techniques like; calculate the mean intensity between images and measure the difference. But are there other more advanced techniques? I know SIFT finds keypoint descriptors in an image so I guess I could compute SIFT descriptors for each image then calculate the number of matches between them. Could I use an integral image here, histogram backprojection?

In computer vision, what are technique/algorithms/ways to measure the homogeneity between 2 images? By homogeneity I mean measure in terms of:

• Colour/intensity for single channel
• Edginess (is that the term), ie, how rough or smooth it is.

As a practical usecase: I have an image that I have divided into grid cells. I wish to select a specific cell and search its neighbourhood (8 cells around it) and identify which of those neighbouring cells are most similar (similar homogeneity) to the subject cell.

I am aware of simple techniques like; calculate the mean intensity between images and measure the difference. But are there other more advanced techniques? I know SIFT finds keypoint descriptors in an image so I guess I could compute SIFT descriptors for each image then calculate the number of matches between them. Could I use an integral image here, histogram backprojection?