Enhanced decoding of YJK images

Pagina 4/5
1 | 2 | 3 | | 5

Van enribar

Paragon (1113)

afbeelding van enribar

09-11-2015, 14:50

Ok now I understand your algorithm, it uses the horizontal shift with 2 images.
I will test the quality on my real MSX2+.
But I think that the images I suggested to you, don't use shifts or any other video "refresh-trick".
About the packages: it's a compilation of files sorted by machine type, so look inside the MSX2+ section.
Each folder or subfolder is a floppy disk, so specify the folder in openMSX or blueMSX as a floppy directory and then start emulator. Programs will start authomatically since most of them have autoexec.bas or .bat, and will show photo demonstrations, slideshows, etc.
I cannot say if there are "modern" tools to show old S12 or SCC images, I remember some of them were displayer with ViewMSX and others.

Van hit9918

Prophet (2901)

afbeelding van hit9918

10-11-2015, 23:29

it made every second line different.
but the pictures were noisy anyways.
with clear pictures the lines pattern would disturb more.

Van KdL

Paragon (1387)

afbeelding van KdL

21-08-2017, 17:01

Here a simple Lanczos3 Pictures Disk

Van sd_snatcher

Prophet (3450)

afbeelding van sd_snatcher

22-08-2017, 16:27

@KdL

It's a nice idea to reduce the color bleed. But due to the lack of vsync, it's causing tearing in the resulting image. You'll probably will have to use XBASIC to make it fast enough.

If you add interlace to the mix, the results would be even more impressive.

Van sd_snatcher

Prophet (3450)

afbeelding van sd_snatcher

23-08-2017, 00:12

Hey, on further inspection even for the individual images (without the flickering trick) you have done quite a feat here!

This is the Ys image you converted:

As a comparison, this is the same image converted with MIFui:

Just ignore the slightly different aspect ratios. And both were done without any dithering.

But the important part is that you seem to have used a much more aggressive Y tweaking. The results on the clouds at the right side are really impressive when compared with the results from MIFui. And MIFui show way more artifacts everywhere.

Great work! Do you mind to share the details of your algorithm?

Van Grauw

Ascended (9900)

afbeelding van Grauw

23-08-2017, 02:10

Yes, I’m also quite interested in this! Big smile

Thinking about using screen 11 at some point, and if so, to write a screen 11 converter for my toolchain, would be good to know what JK plane selection methods give high quality. (Hit’s idea a few pages back was also nice.)

Van Sander

Founder (1865)

afbeelding van Sander

24-08-2017, 00:04

Great conversion KDL!

Van Louthrax

Prophet (2404)

afbeelding van Louthrax

24-08-2017, 01:23

Amazing! I'm really curious about the details too! (snatcher, where did you get that image?)

Van sd_snatcher

Prophet (3450)

afbeelding van sd_snatcher

24-08-2017, 02:12

@Louthrax

Check your e-mail. Wink

Van hit9918

Prophet (2901)

afbeelding van hit9918

24-08-2017, 03:29

I thought KdL used my converter. There is a link some pages ago, with source.
I used the VDPs YJK->RGB formula to search the right Y. after making JK from a mix of the 4 pixels.

@grauw, yes there might be some more secrets in JK selection.
if one imagines 4 pixels of same Y, maybe sliding JK can give very fine brightness changes.
a game tile could slide in screen 2 linecolor style, with very subtle changes.
but how to bring this together with the normal converter algorithm, mhm.
after the four Y are calculated, loop J and K in +- 4 steps or so. with a brightness rating function.

mhm ok there is another JK selection issue, maybe you meant that. I made JK with the standard formula from the mixcolor. but the 4 pixels may be off the mix color, maybe sometimes another JK is better, mhm.
the problem is that a dumb loop to try out every combination would be about 32 bits, 2^32 loops.

Pagina 4/5
1 | 2 | 3 | | 5