You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 22: Image Processing (48)
john-b-yang

Here's a paper published by several European researchers describing a nonlocal algorithm for image denoising linked here (https://bit.ly/2Il4uQr). The paper first compares non local denoising to existing strategies like local smoothing filters, then applies denoising to a couple examples to demonstrate the differences between the non local means algorithm and local smoothing more visually.

rpottle

Couldn't assuming that repeating structures exist in an image cause a high probability of averaging windows that look similar on a micro scale but are actually from unrelated objects in the overall image?

GitMerlin

I understand that we are taking a weighted average of pixels which have a similar neighborhood as p, but why do we choose the dotted red box as the search region?

anup-h

I'm pretty sure the search region around p is arbitrary; I don't think the dotted red line in the figure is determined in some way from p's position

You must be enrolled in the course to comment