Cellular Areas by Rachel Duim

Cellular Areas by Rachel Duim
http://www.filterforge.com/filters/15396.html

WOW! Awesome, Finally it is available and looks very good, will test it how it works and what can be done

Thanks really very much for taking the time and work on this filter and is much appreciated

I have just seen it inside as was very curious, and I inew that it would not be using the bomber, and is clever how is done.

Will make some tests and see the result

Congratulations that you have you have been able to make a filter with variable size of cells according to the lightness

Thanks again

I like this and is very good and well done, but for some things is not exactly what I expected, because on the more darker side there is too much overlay and tiny mix of polygons and is not clean

I have been trying to understand and fix this but yet have not understood and find out how to solve this

Please, Could you be so kind to explain how the filter works briefly?

It seems that all is made by modifying the Details and Roughness of the Cells component and mix them.

Would be possible to be able to have a less tangled and mix of much tiny polygons on the more darker sides?

Hello, I think it is not your fault as you have done it the best you have thought and is really great and you deserve the Editor´s Pick for this awesome filter

So as I really like much this and is very well done, is a great pity that it is not giving better results, so I have been making some tests using different values for the Cells components and seeing how it works and have discovered 2 things

1 - DETAILS VALUES ON CELLS COMPONENTS
The Details values should be lowered and be= 0, 5, 10, 20, 25 to make it better. These values could be maybe made even better and optimized with more tests and adjust them even more so it maybe could give even more gradual and increasing effect, and will do it later.

It seems that the Roughness values are not important, because changing them does not make any difference

2 - EDGE DETECTOR COMPONENT
It seems that having only one Edge detector makes it difficult to capture all the fine details, So I have seen that the Edge Detector component should be separated and have 5 of them, one for each of the steps, and so this way it gives a much more clean and much better polygons without mixing them and with more detail as I have tested

The tests are done connecting the profile gradient shown below as the source imageso it give a fast and very good way to test it

So this below is how I have done it

And here is the result of the original default filter, as you can see the above one is much clearer and better, although the progressive difference size of the polygons may not be as good, so I will try modifying again the details values to optimize them.

Feel free to upload the test example here. Unfortunately my experience with the details and roughness was that it wasn't linear but... I'd like to see what you did. Having 5 edge detectors will slow it down a lot, that's why I did not do it that way. I will consider making an update if it can be improved.