Frequency Domain Image Compression and Filtering

blog.demofox.org

62 points·vonadz··6 comments

I enjoy reading this author's articles such as the one explaining auto-differentiation [1]. Before that, I tried to understand this survey on auto-differentiation [2], and it just didn't click for me. The analogy to imaginary numbers was perfect.

[1] https://blog.demofox.org/2014/12/30/dual-numbers-automatic-d...

[2] https://arxiv.org/abs/1502.05767v4

Great article. The author also links to Steve Brunton's video on A Compressed Overview of Sparsity. I found Mr. Brunton's channel only recently myself and found it to be a treasure trove of information. Highly reccomended.

https://www.youtube.com/watch?v=aHCyHbRIz44&feature=youtu.be

There was an article that demonstrated scratch removal by zeroing out (or was it blurring?) the its dual in the DCT domain. Unfortunately I can't dig it up.

Here are a few resources for periodic pattern removal (e.g. in halftone print images) by zeroing out parts of the image in the Fourier domain. Maybe that was what you had in mind?

http://matzjb.se/2015/08/08/smoothing-a-halftone-photo-using...

http://scipy-lectures.org/intro/scipy/auto_examples/solution...

https://developer.apple.com/documentation/accelerate/halfton...

https://stackoverflow.com/questions/29235421/find-proper-not...

https://stackoverflow.com/questions/34027840/removing-period...

https://www.youtube.com/watch?v=yyox358zIRw

I did exactly that in a grad level image processing take home final and got a 70. There were other corruptions in the image. The professor never taught anything about image correction or evaluating error, he wanted to see what I came up with. I still don’t understand that grade.

Found it[1]. I misremembered a bit, they demonstrated how to remove periodic noise, then left "remove the string" as an exercise for the reader (bottom of the page).

[1] https://legacy.imagemagick.org/Usage/fourier/

Reply on news.ycombinator.com