A new way to store raster images in flash

Discuss the future, present and past of sequential art.

Moderator: Moderators

Locked
ShadowCaster
Frequent Poster
Posts: 93
Joined: Tue Sep 23, 2003 10:07 am
Location: France

A new way to store raster images in flash

Post by ShadowCaster »

I know I have way too much projects at the same time (fractal spiral on hold until someone finds me a circular story BTW) but I think I'm on something original and potentially very usefull in conjunction with the tarquin engine :

The tarquin engine works with embedded raster images. These images can make the flash swf files quite big.
But flash has built in vetorial support. My idea is to store in the flash the images as a combination of vectorial and raster data :
Each image would be vectorized in a LOSSY way that would make most pixels correct when rendered at the pixel dimension of the original image, and an almost fully transparent raster image would hold the data for the pixels wrongly rendered with the vectorial encoding. Then encoder would try different levels of vectorial encoding, trying to optimize the size of Vector data+Raster data.
For the reader, it would mean :
- a pixel perfect reading experience, compared to the original raster based encoding.
- far smaller .swf files for webcomics like Mimi's last coffe that work incredibly well at being reencoded with this techinque (because each pannel is made with a few big monochromatic shapes, witch is precisely the kind of data that can be encoded in a very compact form with vector data)
- a bit of precomputing to regenerate the raster data form the Vectorial+raster. Ideally, this shoud be done during the download of the images for that time to remain hidden. (but it would show in case of offline reading)
Merlin
Frequent Poster
Posts: 81
Joined: Wed May 30, 2001 7:00 pm
Location: England
Contact:

Post by Merlin »

This is a bit beyond my level of understanding but, if you figure out a way of doing it, it sounds like a handy idea.

The problem from the engine's point of view would be that the images would absolutely need to be purely raster based when the comic is actually being read - rasters resize quickly and smoothly in Flash even on slower systems, but vectors eat up resources and get all jerky (especially when there's a whole comic full of them).
New Experiments In Fiction

www.E-merl.com
ShadowCaster
Frequent Poster
Posts: 93
Joined: Tue Sep 23, 2003 10:07 am
Location: France

Post by ShadowCaster »

I agree with the problem you point out that the images manipulated by the engine should be 100% raster for speed performance. This is why some precomputing to build raster images from Vectorial + Raster is needed. I haven't got a working version of this yet but it should be doable. If not, keeping that technique while taking vectorial rendering speed into account in the encoding process from raster to V+R should be tested. Well placed raster pixels can hide the jerkyness of the vectorial rendering (BTW this is usually due to the engine rendering in low quality to comply with the speed requirements of the animations. Offscreen raster images are delt in no time at all because the computation of the intersection of the viewport and those images is just a matter of testing four points, and raster image zooming is rendered so fast that I wish I knew how the Flash team implemented it, maybe some hardware video card access in the plug-in?).
I just have 1 week left before vacation but will seriously work on that problem in September.
Locked