Interpolation by Optimization

asked 2014-04-14 19:45:49 -0600

Royi gravatar image

updated 2014-04-14 19:50:04 -0600

I have the following interpolation problem:

Given an M by N channel image (1 Channel).
A certain set of pixels are defined as anchors. The rest of the pixels should be interpolated using the original data by minimizing the follwoing cost function:

image description

Where E is the new matrix of interpolated pixels and I is the original image.
Pay attention that the interpolated image equals the original image at the anchor pixels.

The inteprolation weights are given by:

image description

My question is, how can I formalize it in Weighted LS form?
Any optimization could be made to solve it quickly and efficiently?

This interpolation could be used as a minimization for many application in image processing (Edge Preserving Interpolation, Colorization, etc...).

Thank You.

edit retag flag offensive close merge delete


Anyone? It should be pretty widely used optimization problem.

Royi gravatar imageRoyi ( 2014-04-15 09:44:52 -0600 )edit