Quote:
Originally Posted by cacapee
Yeah, I looked into the rasterfarian code (old and new) when writing my tool. As far as image processing goes, I'm using CImg (simply because it is easy to integrate). The process goes like this
generate images at 300 dpi
dilate (using CImg)
resize (using Cimg's cubic interpolation)
convert to gif using (libgd)
I believe you are correct in that things can be improved. I was looking at using the lanczos filter to do resizing. Also remapping to the 4-color palette as opposed to using whatever libgd does may be a better thing to do. I'll also add sharpen.
I want to get a stable code base before releasing the source. If you are interested I can add a output-debug-image option that will save out bmps at each step of processing. This is probably more useful to you too.
|
When I developped my
manga2ebook tool I've did a comparison on all available filters of imagemagick too. In a nitpick test I find the catrom-filter to produce the best quality rasterized grayscale images. The lanczos is a good 2nd place though.
You might want to use catrom instead nevertheless.
One of the best thing of rasterfarian is that it uses multiple processes so it can make optimal use of the available processor. That might a be a good idea for you too?
But the reason I'm sticking with rasterfarian must be the TOC-feature. Unfortunately, that is the thing that uses the pdftools which caused the whole ruckus...
Also.. converting the images to png and then using some png-optimizer could give even better result.