DPX files

[Dan Hudgins]

DPX > XnView > DANCAD87.EXE (tm) > BMP > JPG

DPX > XnView > DANCAD87.EXE (tm) > BMP > JPG

Here you go

This is RAW data converted to DPX 10bit log format using Camera RGB and RedLog curves to maximise the log information.

Paul

Thanks for the sample DPX direct form the RED ONE (tm)!

I seem to have been able to use XnView to do conversion without any gaps in the histogram so I do not see 8 bit damage to the tone values, but that may be due to my workflow.

I have attached some histograms and sample conversions,

DDPXL1A.JPG = Converted data with white and black levels pushed in, notice that the histogram does not fill the 16 bits for each primary, so XnView did not clip the ends of the histogram off.

DDPX01A.JPG = Sample grading with just white and black level adjusted.

DDPX01H.JPG = Histograms showing no gaps so far.

DDPX02A.JPG = Sample grading from DDPX01A with some curves and color matrix to bring out more in the dogs eyes. If the shot is only 2 seconds long it can help to have the eyes "pop" better. I could have fiddled with the gray balance but left it as it was.

DDPX02H.JPG = Still do not see gaps in the Histogram display, so far, so the conversion seems to work. Nice to have the 4K DPX as a source for testing, thanks.

It would be good to know if there is any difference between the 48bit tiff output and the DPX that can be seen, if you could post the 48bit tiff of that same frame we could compare the results(?) The attached images are for viewing at video gamma 2.4, the screen shots have a two black and two white ref patches near the gamma 2.4 ref in the lower right, you may have problems seeing the edge between the near white and the white patches on a LCD monitor, but on my CRT I can see the difference, so if the whites look clipped check the white ref patches to see if it is your monitor. The black ref patches are just above the black background, and somewhat above the balck background, you should be able to just see both if the monitor brightness is set right. The gamma patch is for un-compensated CRT monitor or TV, if it does not look right you can load the images into a graphics program and adjust the monitor gamma comp in that program until the horazontal lines in the small gray box in the lower right looks the same as the gray box around it when you squint from a few feet away form your monitor.

 

[Dan Hudgins]

What program was it graded in?

What program was it graded in?

Graded Version :innocent:

What program was it graded in?

What settings did you use, sat, gamma, etc.?

 

[Paul Leeming]

Here you go:

Full Size TIFF (RedLog) - 48MB

Full Size TIFF (Rec.709) - 48MB

Full Size TIFF (colour corrected in RedCine using Rec.709) - 48MB

...and my colour correction as a 25% JPG (done in RedCine) so you can compare it here without downloading lots of TIFFs!

Cheers,

Paul

 

[RedFriend]

Very cool!

 

[Dan Hudgins]

Thanks for making sample TIFF frames also...

Thanks for making sample TIFF frames also...

Here you go:

Full Size TIFF (RedLog) - 48MB

Full Size TIFF (Rec.709) - 48MB

Full Size TIFF (colour corrected in RedCine using Rec.709) - 48MB

...and my colour correction as a 25% JPG (done in RedCine) so you can compare it here without downloading lots of TIFFs!

Cheers,

Paul

Thanks Paul for making the sample RED ONE (tm) DPX and TIFF files available for testing, your time spent in doing this is appreciated very much.

The subject of this frame is also interesting since it will be a hard image to hold up in a film out do to the need to get some tones with contrast at both ends of the scale. A subject that is colored near middle gray illuminated by strong side light gives a tonal range do to the shadow, whereas this image the tone range is mostly do to the coloring of the subject due to the flat lighting, so the contast in the dogs face and the bottle neck label are of low contrast, if you try to get wide contrast in the face it is hard to hold the label, and if you try to get wide contrast in the bottle neck label it is hard to hold the contrast in the dogs face. If this had been shot under studio lighting, one could put a spot light on the dogs face, rim light, and cut the light on the bottle label to rase the contrast on each subject while decreasing the brightness difference between the two of the subjects in the frame.

Not to be taken as a criticism, but looking at your grading example on my brother's computer's PC CRT monitor the dog becomes very dark with little detail, I showed him my senond timing where I tried to bring out the brown in the dog's eyes without losing the bottle neck label's pattern, and he thought that it was more visable over all. There seems to be a trend for people on RED USER (tm) to grade their frames from too dark to very much too dark, maybe because they are using LCD monitors that tend to have poor highlight detail, I have noticed that my LaCie CRT monitor seems to give better highlights than my LCD monitor.

The Monitor gamma test patterns one can download sometimes have a checker board 50/50% patch which can give a false monitor Gamma reading, the charts with horizontal lines of 50/50% probably work better, because of monotor peaking or sharpness boost on the checkerboard pixels changing their brightness value.

I would recomend for people doing their timing to make test shots and burn a DVD to look at on various CRT, LCD, DLP, OLED, etc. TV sets to check that they are not making their grading too dark or light in the important parts. Many people watch TV with the room lights on or with daylight in the room. For doing a filmout making a test strip of key frames from each shot on a film print may be better than trying to rely on a 3D LUT and monitor, I have a test strip option in my Aggregate command in my Edit list. Since both Dupe-Neg and Print films have an S shaped curve as you move away from mid-tone the contrast decreases, that is you get less contrast in the highlights and shadows than in the mid tones, so filmmakers tried to light to hold the subject in the film's mid-tones and grading lights were adjusted to keep the prime subject in the mid-tones, while other parts of the frame came out softer in contrast. With digital the system has hard clip points even if you apply a S curve, older negative and dup film stocks had no real sudden hard top or bottom if pull processed at a good lab, the contrast just fell lower and lower at both ends and got grany. Newer film stocks have been altered to have a sharper toe to get a "good black" but that can be adjusted with less bromide use.

Anyway, thank you for your generosity!

 

[M Most]

There seems to be a trend for people on RED USER (tm) to grade their frames from too dark to very much too dark, maybe because they are using LCD monitors that tend to have poor highlight detail

More likely it's because they're using Macs with native 1.8 gamma.

 

[Paul Leeming]

Not to be taken as a criticism, but looking at your grading example on my brother's computer's PC CRT monitor the dog becomes very dark with little detail, I showed him my senond timing where I tried to bring out the brown in the dog's eyes without losing the bottle neck label's pattern, and he thought that it was more visable over all. There seems to be a trend for people on RED USER (tm) to grade their frames from too dark to very much too dark, maybe because they are using LCD monitors that tend to have poor highlight detail, I have noticed that my LaCie CRT monitor seems to give better highlights than my LCD monitor.

No criticism taken. I use an Eizo S2411W LCD monitor calibrated with the ColorVision Spyder2 so as far as LCDs go it's among the very best to grade on.

The graded version has no clipping in either highlights or shadows so if the dog (named "Lin" btw, not mine!) looks too black then it might be something in the monitor you're using, like the 1.8 gamma someone suggested? No idea but given you're messing with high level colour profiling stuff I suspect your monitors are up to the task, so who knows?

Also I do have the advantage of knowing the dog and her fur IS very jet black, almost velvety, so I did my grade to accurately represent how she looks and not clip any detail (which is there if you bump up the gamma).

But yes, I chose that shot for you based on the wide contrast and lighting to make it challenging! Plus dogs don't ask for royalties....

Cheers!

Paul

 

[Peace Villow]

What program was it graded in?

What settings did you use, sat, gamma, etc.?

I used Apple Color.
Nothing special in there, just balancing the Lift, Gamma, and Gain.
Pull a key on the green bars and the blue jeans.
Add custom shape to darken the right side, and finally add a vignette around the dog.

Not to be taken as a criticism, but looking at your grading example on my brother's computer's PC CRT monitor the dog becomes very dark with little detail, I showed him my senond timing where I tried to bring out the brown in the dog's eyes without losing the bottle neck label's pattern, and he thought that it was more visable over all. There seems to be a trend for people on RED USER (tm) to grade their frames from too dark to very much too dark, maybe because they are using LCD monitors that tend to have poor highlight detail, I have noticed that my LaCie CRT monitor seems to give better highlights than my LCD monitor.

How about my version?

I use uncalibrated Apple Cinema Display, it look dark.
But when I see in my broadcast monitor (via decklink extreme) it look well balanced.
When I upload this image with my PC at home it look slightly dark.

 

[Dan Hudgins]

Looks good!

Looks good!

I used Apple Color.
Nothing special in there, just balancing the Lift, Gamma, and Gain.
Pull a key on the green bars and the blue jeans.
Add custom shape to darken the right side, and finally add a vignette around the dog.

How about my version?

I use uncalibrated Apple Cinema Display, it look dark.
But when I see in my broadcast monitor (via decklink extreme) it look well balanced.
When I upload this image with my PC at home it look slightly dark.

This looks good, maybe you could get a little more pink in the bottle label and brown in the dogs eyes, but I think the vignette helps a lot, I have been thinking of adding something like that when I have the time. I was going to mention that Ansel Adams sarted the f/64 group he was into "straight" printing on low contrast paper, but later used higher contrast paper and "Dodging and burning" to give both light and dark subjects in a frame more contrast.

Hard to use Dodging and burning on shots that the camera pans or moves though.

You image looks good on my LaCie gamma 2.4 CRT monitor.

I was thinking that maybe REDUSER.net should adopt a gamma standard for images that are uploaded (how they display changing the gamma on the fly while uploading), when you upload an image you could type in the gamma the image was adjusted for, and it would bring it to 2.4 or the standard for internet images on an uncompendated monitor (is there a standard for internet images? is there a gamma standard for DVD images? are those the same?)

When you can light then use spot lights on dark parts and less light on bright parts, when you cannot like in a landscape then Dodging and burning seems a valid way to get things at the ends into center with better contrast, thanks for the example made with real RED data so we can see the dark nose and such!

 

[Daniel García]

I was thinking that maybe REDUSER.net should adopt a gamma standard for images that are uploaded (how they display changing the gamma on the fly while uploading), when you upload an image you could type in the gamma the image was adjusted for, and it would bring it to 2.4 or the standard for internet images on an uncompendated monitor (is there a standard for internet images? is there a gamma standard for DVD images? are those the same?)

Or maybe use a color managed browser, such as safari, or (IIRC) Firefox 3?

 

[Peace Villow]

Hard to use Dodging and burning on shots that the camera pans or moves though.

It's easy if you're using Lustre or Resolve :biggrin:
I think Lustre have the best tracking system, you can use point tracker or region tracker.
And you can assign multi shape/geometry under the same axis point.

 

[Peace Villow]

Or maybe use a color managed browser, such as safari, or (IIRC) Firefox 3?

I don't see any difference when using IE/Firefox/Safari side by side.