Login at Kodi Home

Sunflux · (This post was last modified: 2015-09-29, 10:02 by Sunflux.)

I just tested the EGL build with 16-235 + Kodi Full + Dithering + TV Limited... and it's nowhere near as clean as 0-255 + Kodi Limited + TV Limited -or- 0-255 + Kodi Full + TV Full. Both of those two present flawlessly smooth grayscales, however in 16-235 mode I can always see banding. It's far less obvious with dithering enabled, but still there.

To be honest, I think what I'm seeing are slight inconsistencies in the dithering pattern that are being amplified when blown up to 4k resolution on my TV. I took a screenshot and, on my computer monitor, the pattern that looked obviously bad on the TV actually looks quite smooth. At least, it does until I zoom in close and scroll side-to-side, where I can make out inconsistencies that form similar banding patterns.

Hufvudet · 2015-09-29, 09:57

(2015-09-29, 09:55)Sunflux Wrote: I just tested the EGL build with 16-235 + Kodi Full + Dithering + TV Limited... and it's nowhere near as clean as 0-255 + Kodi Limited + TV Limited -or- 0-255 + Kodi Full + TV Full. Both of those two present flawlessly smooth grayscales, however in 16-235 mode I can always see banding. It's less obvious with dithering enabled, but still there.

Thanks for testing all this and the charts btw!

**fritsch** · 2015-09-29, 10:03

(2015-09-29, 09:55)Sunflux Wrote: I just tested the EGL build with 16-235 + Kodi Full + Dithering + TV Limited... and it's nowhere near as clean as 0-255 + Kodi Limited + TV Limited -or- 0-255 + Kodi Full + TV Full. Both of those two present flawlessly smooth grayscales, however in 16-235 mode I can always see banding. It's far less obvious with dithering enabled, but still there.

To be honest, I think what I'm seeing are slight inconsistencies in the dithering pattern that are being amplified when blown up to 4k resolution on my TV. I took a screenshot and, on my computer monitor, the pattern that looked obviously bad on the TV actually looks quite smooth. At least, it does until I zoom in close and scroll side-to-side, where I can make out inconsistencies that form similar banding patterns.

Good - that's what I expected :-) At least we now have a "better out of the box experience" for everyday joe.

Hufvudet · 2015-09-29, 10:26

And btw nr. 2 @Sunflux. While you're at it. Did you test with dithering enabled/disabled when going:

Driver 0-255 -> Kodi Limited -> TV Limited -> CORRECT (Best)

I know we say dithering should help with bad rips/banding in source but do your tests reveal any difference?

**fritsch** · 2015-09-29, 10:39

We first need to find such a bad rip - perhaps the sample from last time is a candidate?

Sunflux · (This post was last modified: 2015-09-29, 10:58 by Sunflux.)

I just completed a bunch more tests.

I'm wrong about the dithering being the cause of the remaining banding I'm seeing in 16-235 mode. I took a series of screenshots, and the screenshots (and dithering) applied in 0-255 + Kodi Full + TV Full (which shows no banding), is identical to the screenshot I get from 16-235 + Kodi Full + TV Limited. So, I think the banding is being introduced further down the line - perhaps by the video driver? It could be the TV, but it's extracting the same color region in both examples, so I don't think that's it.

Ideal modes:

0-255 + Kodi Full + TV Full + Dithering On = perfectly smooth but heavily dithered image
0-255 + Kodi Limited + TV Limited + Dithering On or Off = perfectly smooth image*

* I cannot see any visual difference between dithering on or off, however the non-dithered screenshot is 27kb, and the dithered screenshot is 33kb, so something is happening. Also, the heavily dithered screenshot is 494kb, so you can see how much it's doing. I think if there's going to be any actual difference made to a bad rip, you're going to have to be in 0-255/Full/Full mode, since that's already heavily processed all around.

Hufvudet · (This post was last modified: 2015-09-29, 10:59 by Hufvudet.)

And if you try this with 0-255 + Kodi Limited + TV Limited: http://workupload.com/file/2KEkdaeY (from Divergent)

Any visual difference at all with dithering on/off?

The banding is scary in that sample so this might be a TOO bad sample. Ideal would maybe be a sample with minor banding in source or rip.

Sunflux · (This post was last modified: 2015-09-29, 11:18 by Sunflux.)

Sorry, there's absolutely no change to the banding in 0-255 + Kodi Limited + TV Limited whether dithering is on of off. However, there is a modest improvement in 0-255 + Kodi Full + TV Full.

Hufvudet · 2015-09-29, 11:23

Do you have any other movie/video with a lot less, but still some proven banding in the source? Divergent is so far out that nothing helps.

And thanks again for testing.

Sunflux · (This post was last modified: 2015-09-29, 11:31 by Sunflux.)

The thing is, I don't think the dithering filter is being applied as a general smoothing - it's being applied specifically when the dynamic range is being changed. In 0-255 + Limited + Limited, no change to the original video is occurring, so there's nothing to dither. If it were going to dither anything, it would be my greyscale test (which is very much dithered in 0-255+Full+Full), but there's essentially no visual difference.

Hufvudet · (This post was last modified: 2015-09-29, 11:28 by Hufvudet.)

Perhaps one of these: https://www.google.se/?gfe_rd=cr&ei=NFMK...or+banding

Hufvudet · 2015-09-29, 11:30

(2015-09-29, 11:28)Sunflux Wrote: The thing is, I don't think the dithering filter is being applied as a general smoothing - it's being applied specifically when the dynamic range is being changed. In 0-255 + Limited + Limited, no change to the original video is occurring, so there's nothing to dither. If were going to dither anything, it would be my greyscale (which is very much dithered in 0-255+Full+Full), but there's essentially no visual difference.

Then that's interesting! @fritsch?

Sunflux · (This post was last modified: 2015-09-29, 11:40 by Sunflux.)

Here's an enlarged example of the greyscale test. The top image is 255+Full+Full+Dithering (dynamic range expanded). The bottom image is 255+Limited+Limited+Dithering (dynamic range untouched). Note that I've exaggerated the contrast to make the dithering more obvious. Dithering is clearly not doing anything in the bottom example, and I'm inclined to believe that is the intended behavior.

Hufvudet · (This post was last modified: 2015-09-29, 11:59 by Hufvudet.)

Quote from madshi / madvr guru:

---
It is true that with a prefect color neutral grayscale, letting madVR render to 16-235 means that madVR doesn't have to dither. However, even black & white movies are usually not 100% color neutral. They often have a slight color tint. And as soon as chroma is anything but exactly color neutral, madVR instantly has to dither. Which is the case for 99.99% of all movies out there.

/unquote
---

Could it be that a grayscale test isn't optimal for testing dithering in 255-limited-limited? The above reasoning would be in line with your grayscale test. That's why it would be interesting to see results with real source with minor banding inherent (not perfect grayscale).

Correct me if I'm wrong.

**fritsch** · 2015-09-29, 11:57

We will find out :-) I asked Lauri to answer the post above in detail. Let's wait a bit. From the Code it is applied - but we will see and get information from the author.