It's getting there! I still think the next major thing is to use dynamic ordering (if a pixel has had its darkness reduced, it should be treated later).
The cat result looks quite good right now, but Lena doesn't, especially in the gradient areas.
Once this works then we can work on different functions for deciding how much darkness to take. The functions can depend on local information (e.g., on a LAHE of the original image) and on the distance along the spiral (take more at the beginning, less as you get further from the black pixel).
Finally, we can get back to stippling, with black dots bigger than one pixel, and treat the halftoning part by having the target darkness correspond to the size of the stipple.
The reason for the method works for preserving the edges is because like what you said the re-ordering will reduce the clumping but in order to keep the edges the clumping will produce the line-like effects, which is good for textures like cat or tree images.
There is one problem for choosing pixels in the almost-white area. I have to find out where the almost-white area is, which might have no pixels chosen. But right now there is still some pixels chosen since the spiral attains its limit. Also it is very slow when doing the almost-white area because the size of spiral is very large and it computes too much pixels.
It's getting there! I still think the next major thing is to use dynamic ordering (if a pixel has had its darkness reduced, it should be treated later).
ReplyDeleteThe cat result looks quite good right now, but Lena doesn't, especially in the gradient areas.
Once this works then we can work on different functions for deciding how much darkness to take. The functions can depend on local information (e.g., on a LAHE of the original image) and on the distance along the spiral (take more at the beginning, less as you get further from the black pixel).
Finally, we can get back to stippling, with black dots bigger than one pixel, and treat the halftoning part by having the target darkness correspond to the size of the stipple.
The reason for the method works for preserving the edges is because like what you said the re-ordering will reduce the clumping but in order to keep the edges the clumping will produce the line-like effects, which is good for textures like cat or tree images.
ReplyDeleteThere is one problem for choosing pixels in the almost-white area. I have to find out where the almost-white area is, which might have no pixels chosen. But right now there is still some pixels chosen since the spiral attains its limit. Also it is very slow when doing the almost-white area because the size of spiral is very large and it computes too much pixels.