zebra show overexpose better when rgb zebra is used - page 3 - General Discussion and Assistance - CHDK Forum supplierdeeply
« previous next »
Pages: 1 2 [3]   Go Down

zebra show overexpose better when rgb zebra is used

  • 28 Replies
  • 14918 Views
*

Offline reyalp

  • ******
  • 14080
Re: zebra show overexpose better when rgb zebra is used
« Reply #20 on: 08 / May / 2011, 15:14:12 »
Advertisements
Quote from: Bernd R on 08 / May / 2011, 15:02:19
brightness 120
saturation 240
tone 80

Is that YUV ?
No, that would be HSV http://en.wikipedia.org/wiki/HSV_color_space

YUV http://en.wikipedia.org/wiki/YUV (see also http://en.wikipedia.org/wiki/YCbCr )

Quote
If so, then if zebra check only y channel, a overexpose can not detect, because brightness is only 120
You are making assumptions, the image has undergone many transformations by the time it gets to MS paint ;)
Logged
Don't forget what the H stands for.
*

Offline Lebeau

  • ***
  • 187
Re: zebra show overexpose better when rgb zebra is used
« Reply #21 on: 08 / May / 2011, 18:37:22 »
What means zebra, or under/over exposure ?

It checks floor and ceiling limits relatively to a color space. If color space is:
 - YCbCr (or YUV, likes), Y means luminance and floor/ceiling over CbCr is really interpretative;
 - RGB, each tonal component represents additive color model's components. Each includes luminance and saturation of these components. Floor/Ceiling limits, apply to each or combine components are relevant to the luminance/saturation within this tonal component;
 - HSL, Hue/Saturation/Lightness, it has the same complexity of interpretation as YCbCr where L is lightness and Hue/Saturation represent tonal and intensity of this tonal component.
 - CieLab for LAB Cie color space where L means Lightness and a/b have similar interpretation as UV
 - CieXYZ, ...
 - ...

Well, the principal question is "What do you want to detect ?", "What do you want to check against what ?"

See, when I look at "Y" zebra on my display, it signals that the luminance is under/over predefined limits. When I set it to something else, I shall interpret zebra upon the selected color space component.

For example, supposed I developed CieLab zebras. "L" refers to lightness similarly to "Y", and a/b refer to signed color components (-127...+127) where I could put -100/+100 as counter-floor/ceiling. But I still have to interpret the meaning of these zebras.

For conclusion, RGB's R G B zebra, like CieLab's a/b, or YCbCr's Cb/Cr, or other color space's color components, are subject to "perceptual/personal" interpretation.

IMHO, the main purpose of zebra, and histogram too, is to display mathematical result from YCbCr viewport datas, refering to spectral information.
Logged
*

Offline reyalp

  • ******
  • 14080
Re: zebra show overexpose better when rgb zebra is used
« Reply #22 on: 08 / May / 2011, 19:12:09 »
It seems to me that as a user trying to take raw pictures, you want to know which pixels in the raw will be >= white level or <= black level, or within some arbitrary margin of these.

If you are taking jpeg, the above may be sufficient as well, but that depends on Canons jpeg process.

Any attempt to get this from the viewport data is going to be an approximation. It's not clear to me that you can determine the best approximation without careful experiment. To the best of my knowledge, the transformations form sensor data->viewport YUV and sensor data->jpeg are not well characterized.
Logged
Don't forget what the H stands for.
*

Offline Bernd R

  • ****
  • 430
Re: zebra show overexpose better when rgb zebra is used
« Reply #23 on: 09 / May / 2011, 07:29:32 »
Quote from: Lebeau on 08 / May / 2011, 18:37:22
What means zebra, or under/over exposure ?

Well, the principal question is "What do you want to detect ?", "What do you want to check against what ?"

I want detect if Camera does not overexpose the important objects in red green and blue chaannel.white overexpose is handle with default zebra good.But if you want see if colors are overexpose you need switch to rgb zebra.

So i see default zebra is not usefull

I have find a converter tool that show with a given rgb value the YUV Output

http://web.forret.com/tools/color.asp?R=255&G=80&B=40

I test
R 255
G 80
B 40

Y is only 50%

so if red channel is overexpose can not detect with Y only.

when use

R 255
G 255
B 40

http://web.forret.com/tools/color.asp?R=255&G=255&B=40

Y is 90%

90% is a value of 255/100 *90 = 229

So zebra is only show with that color, if overexpose threshhold is set to 26.but then white overexpose is show too early

above colors i get in real world shoots

« Last Edit: 09 / May / 2011, 07:32:41 by Bernd R »
Logged
Ixus 1000 HS
*

Offline Lebeau

  • ***
  • 187
Re: zebra show overexpose better when rgb zebra is used
« Reply #24 on: 09 / May / 2011, 07:45:58 »
I understand your target.

The question is:
Is sensor data available before shooting in order to read preliminary sensor RGGB data before shooting and therefore create a sampled RAW histogram/zebra display ?
Logged
*

Offline Bernd R

  • ****
  • 430
Re: zebra show overexpose better when rgb zebra is used
« Reply #25 on: 09 / May / 2011, 08:05:16 »
Quote from: Lebeau on 09 / May / 2011, 07:45:58
I understand your target.

The question is:
Is sensor data available before shooting in order to read preliminary sensor RGGB data before shooting and therefore create a sampled RAW histogram/zebra display ?

I do some testcode for Ixus 1000 that show frequently 4 pixel values as this

pixel pixel
pixel pixel

value of center Pixels( i use the code to find the bayer pattern it use).the RAW buffer is not continues update, the values are only on shoot correct.other time they are always same
But maybe other can test if this work on other Camera better.here is the code

http://chdk.setepontos.com/index.php?topic=5722.msg60215#msg60215

So there must be for Ixus 1000 another RAW buffer that is used to calculate the preview.maybe this buffer is smaller, because for zebra is not every pixel need.

But if have yuv zebra only is still good enough to be usefull if in the calc u v is add to Y and then check this value for overexpose in standard zebra mode to.
Logged
Ixus 1000 HS
*

Offline reyalp

  • ******
  • 14080
Re: zebra show overexpose better when rgb zebra is used
« Reply #26 on: 09 / May / 2011, 12:08:30 »
Quote from: Lebeau on 09 / May / 2011, 07:45:58
I understand your target.

The question is:
Is sensor data available before shooting in order to read preliminary sensor RGGB data before shooting and therefore create a sampled RAW histogram/zebra display ?
If it is, I don't think anyone has ever found it. I would guess that something down in the hardware creates the YUV live buffer directly.

Thinking about my previous post, the sensor->YUV transformation is not constant. The viewport (and jpeg) both have white balance, the raw data does not. So without additional information, it's impossible to establish an RGB zebra from the viewport that really corresponds to the limits of the raw buffer. Canon "i-constrast" would also confuse things.

If the transformation used for the viewport and jpeg are close (one would hope so), you may still still be able to get a good idea of what will exceed limits in the jpeg, I guess.
Logged
Don't forget what the H stands for.
*

Offline Bernd R

  • ****
  • 430
Re: zebra show overexpose better when rgb zebra is used
« Reply #27 on: 10 / May / 2011, 10:20:30 »
Quote
If it is, I don't think anyone has ever found it. I would guess that something down in the hardware creates the YUV live buffer directly.

Thinking about my previous post, the sensor->YUV transformation is not constant. The viewport (and jpeg) both have white balance, the raw data does not. So without additional information, it's impossible to establish an RGB zebra from the viewport that really corresponds to the limits of the raw buffer. Canon "i-constrast" would also confuse things.

If the transformation used for the viewport and jpeg are close (one would hope so), you may still still be able to get a good idea of what will exceed limits in the jpeg, I guess.

If  RAW zebra is possible, it give correct result when you save as RAW.So if it is possible to access RAW data it should only add when RAW is enable.
when RAW zebra show no overexpose, it happen maybe that jpg is overexpose, but the RAW file is ok.

but even if no RAW zebra is possible when in default zebra mode not only use y and calc r g b same as rgb zebra do, zebra is more precise

usefull for me is too, if there is a after shoot zebra view.that show overexpose in RAW buffer.it is possible before RAW save to generate zebra of RAW for the display.and then during RAW save, this zebra can show.

so time of RAW save can be usefull to look for overexpose, and so photographer can see if he better repeat the shoot.
« Last Edit: 10 / May / 2011, 10:28:01 by Bernd R »
Logged
Ixus 1000 HS
*

Offline Lebeau

  • ***
  • 187
Re: zebra show overexpose better when rgb zebra is used
« Reply #28 on: 14 / May / 2011, 16:05:22 »
Well, I made a little routine to get black and white count for HDR purpose.

http://chdk.setepontos.com/index.php?topic=6379.msg66629#msg66629
Logged
Pages: 1 2 [3]   Go Up
« previous next »
 

Related Topics