I can't use noise since it introduces an uncertainty as to precisely what the pixel value is at any point on the screen, which is contrary to the purpose of an analytical test pattern. The pattern I sent you is intended for visual inspection -- any skipped or repeated values in Y will instantly be visible as an increase of banding, which you can easily see by trying to encode it directly though any YUV encoder. The pattern is designed to sniff out any increased banding imposed by the TV set it is played on, which is why I can't have any skipped or repeated values in the original. For Blu-ray, one is stuck with 8-bit encoding for signals on the disc, which is why I need a way to convert one channel of an RGB signal directly into a Y signal, without any recalculation, just move the selected channel's data into the Y channel. If I knew the precise file format of, say, an 4:2:2 YUV AVI signal, I would do it manually -- by writing a program that puts Y U and V pixel data directly into an artificially created file. But I've had trouble finding a suitably detailed description of that file format and it would be my last resort, due to the length of time it would take to write the software.

D 

On Tue, May 22, 2012 at 4:47 AM, Py Fave <pyfave@gmail.com> wrote:

lets keep this on the list. i am curious of other people's responses

as i see it ,the problem is in the media.
if you can't display more than 8 bit colors  ,
you're stuck.

professional systems use SDI connexions on expensive monitors mostly to get these things  solved.
i don't know exactly in video , but in (digital) cinema , you can keep your image 16 bits on screen .
(dpx or tiff support this, f. i.)


anyway , if it is for color calibration, don't consider that the probe will be aligned on a pixel .
it will give a median value from surrounding pixels , this will smooth out your values .

and if you use the "noise" trick (in fact it is called stochastic distribution or something like this :-)
it will give pretty good values.

if you HAVE to output bluray , it will be 8 bits.

16 bits is not affordable  for consumer market  now .
tell me if i'm wrong please .

"I'd be eternally grateful." hehe
have a nice day

Thanks for responding to my query about gradients. You're the first one to do so.

The MPEG-4 encoders for all Blu-ray require 8-bit video in YUV form, which is one of the primary faults of the system. Using ten bit RGB as input to an 8-bit YUV format won't help matters if the results are still rounded off to 8 bits with the same round-off errors as with 8-bit originals.

I realize adding noise will smooth gradients, but these are test patterns where I need the Blu-ray players to produce the PRECISE values I intend so as to test the gradient response of the SCREENS! 

Attached is the file I've been trying to encode without success. Note on the horizontal axis every pixel value between 16 and 235 appears for precisely the same amount of time (except for some stuff going on near the left edge). With every encoding procedure I've tried, either some pixel values are repeated or are repeated, with consequent increased banding on playback.

Thx again for your attention and if you can figure out how to do this I'd be eternally grateful.

David