Well I'we found a small bug :-)
The screen 2 tiles are rotated 90° vertically, so the 2 colors in 8 pixels apply on vertical 8 points in each 8x8 block...
As soon I correct the problem I'll post new results
I hope your enthusiasm is being picked up, artrag It looks cool! Although I don't completely understand what we're looking at
IT's a video encoder that optimizes the pattern table usage.
IT's idea is based on the utopia demo from dvik and joyrex.
And it makes videos possible on MSX1 machines at a decent framerate, right? Using little data?
Yes! Using the same tileset (3x2 = 6K) you can encode 12 frames at 768bytes each (or less if you use dfferential encoding across frames)
All errors corrected:
site updated!
http://ragozini.googlepages.com/vdpenc
next step: to generate an msx rom that holds video
OK I've spotted another bug in the MSX1 color palette.
The colors are completely wrong...
Stay tuned for the next try
OK site updated!!
http://ragozini.googlepages.com/vdpenc
Now the results are correct: the encoded images are from MSX1 screen2 (at least they should ;))
I know, they look worst than the previous version, but this was due to the fact that the msx 1
palette has only one gray level...
Maybe I should change the staring video in something closer to the MSX1 capabilities...
Does anyone have a better example (of 12-20 frames) to be processed?
Yeah, you need something wtih more colors than just black, grey and red
Maybe taking only the bright of the colors and working with them like they were all grey shades, some good results could be achieved. Of course to see the result you would need to turn off color on yout TV (or there was a way to do it modifying a register?)
I think that dithering is the right solution.
Floyd algorithm allows to compensate errors in the color space
spreading them in the adjacent pixels
The problem is to implement it....
I did some other experiments with a tank... the color issue is always the same...