Record similarity function for images for data clustering

Question

I know how to write a similarity function for data points in Euclidean space (taking a negative minimum quadratic error.) Now, if I want to check clustering algorithms on images, how can I write a similarity function for data points in images? Do I base it on my RGB values ​​or what? And How?

Answer 1

I think we need to clarify some points:

• Are you only going for color? So, take the RGB values ​​for the pixels and apply your metric function (minimize the amount of square error or just calculate the SAD - Sum of Absolute Differences).
• ( )? , , . 2D- ... 3D, (. ).
• 3D- ? (2D- + 1D-) . , , . , 3D- 2D 1D, PCA . , - / .

, , , , .