FixBlackPixels

[lcipolina]

Hi, regarding the "Fix black pixels" function. I guess what bothers you is the pixelation?

IF this is the case, there are 2 options:

1. Change the background and map into a white canvas
2. OpenCV remap. This one is a little bit tricky to use. IF you need it, I can try to implement it, or find anotherone that is less cumbersome.

Let me know what is exactly bothering you and I can try to see how I solved it on my code.

Also, there is an openCV discord server. Where people are very nice and knowledgeable: https://discord.gg/PskueT

[Null3rror]

Hi, thanks for helping me out.
I wanted to fill the missing pixels(non integer ones when recreating the image via ReCreateImage function), I think I have to use an interpolation to fix them.
I tried the OpenCV remap function as you suggested, it was really fast, but it required a lot of change in code.
I would be grateful if I can learn from your code(If it's okay with you), so I can implement a better solution.

[lcipolina]

Hi! I actually learned from your code :)
Then I am working with a visual artist to improve the image output, as it is hard.

We changed your code to use the remap function indeed, and as you said, we refactored the entire code, as you exactly said. The remap is faster and works like a charm, but it requires to think backwards. There are forums and forums about this...

Maybe there is a simple way to implement the "remap" nonsense. I can ask on the Discord server, since there is a lot of knowledgeable and experienced people.

I have created another branch were I've uploaded the latest version of our code. It's not done yet and some comments are off (for example, we mention "polars" but then we never used polars).

It is basically your code: "LOG-ROTATE-EXP" but without the tiling. Also, the r2/r1 = 256 for us, as it is the constant used by Escher in his "Print Gallery" painting.

I also a document with the math we used as well. It's an alternative representation of your math. But I've find it a little bit more intuitive. We don't work with the trigonometric representation of complex, we just use the square coordinates.

The PDF document also explains the constant behind Escher's paining, it's just a bunch of math that I used to explain what's going on behind the painting and why our code is this way.

[Null3rror]

Thanks, I'll check it out

[lcipolina]

Sure, let me know if I can be of help. Most of the code was done by our visual artist, since I only took care of the math portion of the painting. I have an idea of what's going on, but if there is something in the code that I don't know, I can ask him.

I also asked on Discord about any alternative to REMAP, I'll see if anyone replies.

[lcipolina]

Hi, just to say that I've asked around on the Discord channels for OpenCV and people still recommend REMAP for pixel interpolation. that's the best alternative they have.

So, I guess, either you use your method... which seems to be working, or either you implement the remap, which is what we did in our Escher code.