Color balance 6D vs. 60D

[privatebydesign]

Just keeping it accurate!

Hopefully when people see how dark a well exposed image actually is they will start to realize why there is so little information in the dark shadows and that tonality will always be an issue when shadows are raised significantly. Let's get true 16 bit RAW files ASAP!

 

[jrista]

Just keeping it accurate!

Hopefully when people see how dark a well exposed image actually is they will start to realize why there is so little information in the dark shadows and that tonality will always be an issue when shadows are raised significantly. Let's get true 16 bit RAW files ASAP!

No kidding. Astro CCD cameras use true 16-bit data, and they have very good noise characteristics (for manufactuers like QSI and FLI, it's extremely close to a true Gaussian distribution...no banding of any kind, and extremely low hot pixel counts). The data is WONDERFUL, even deep down into the darkest tones. I would LOVE for my DSLR data to look like that.


I usually have to stack a couple hundred RAW files from my 5D III before my astro data starts to get that good.


For the record, I do believe PixInsight loads data into a 32-bit float space. I'm actually not sure how that affects the base rendering. This image was QUITE dark...even for an unmodified RAW. I might be able to get it to load in 14-bit native space...kind of curious how it will look, because some of the highlights were pretty close to clipping in a number of these photos.

 

[memoriaphoto]

So let me see if I get this straight. You claim that all RAW files are pretty much the same in its original state and that there's no such thing as "camera/sensor character". It all boils down to profiles and RAW conversion software? Meaning that if I take a 5D Classic and a 5D Mark III and use a Color Checker Passport in the same environment and then start shooting, the output in Lightroom for example, will be virtually the same (with the exception of the higher res in the 5D3)?

Just a reminder that I was referring to how a sensor and CFA reacts to different light and a steeper CFA has better potential to read the scene more accurately and hence produce "better" colors. Are you saying that this is also BS if you have a custom profile for the converter of your choice?

Now, I know there's no such thing as "color" per se in a RAW file but I would have thought logically that it contains "color instructions" to the software and those instructions are not the same due to different color filters and in-camera processing.

 

[jrista]

So let me see if I get this straight. You claim that all RAW files are pretty much the same in its original state and that there's no such thing as "camera/sensor character". It all boils down to profiles and RAW conversion software? Meaning that if I take a 5D Classic and a 5D Mark III and use a Color Checker Passport in the same environment and then start shooting, the output in Lightroom for example, will be virtually the same (with the exception of the higher res in the 5D3)?

The output in lightroom will be all about how LIGHTROOM processes the data. There are likely to be minor differences due to differences in noise levels and characteristic, chroma noise can impact the fundamental color cast of an image, but overall...what you see as a result of those two images being processed in Lightroom is what LIGHTROOM ITSELF does to the data. Run the same experiment with DPP...you will get obviously different results compared to the Lightroom test, and the magnitude of difference between LR and DPP will be greater than the magnitude of difference between the two cameras. Same goes for using Aperture, or RawThearapee.

It's MATH. Everyone uses different algorithms and different curves when they process the data...so what your seeing on screen has far, far less to do with the camera hardware, and far far more to do with the mathematics that drive these programs.

Just a reminder that I was referring to how a sensor and CFA reacts to different light and a steeper CFA has better potential to read the scene more accurately and hence produce "better" colors. Are you saying that this is also BS if you have a custom profile for the converter of your choice?

A stronger CFA reduces color cross between pixels. That means each pixel represents a purer color. When standard algorithms such as ADH are used with data that came from a sensor with a stronger CFA, color noise should be lower, which will reduce any color cast caused by that color noise. As I said, IQ is more about noise.

Now, I know there's no such thing as "color" per se in a RAW file but I would have thought logically that it contains "color instructions" to the software and those instructions are not the same due to different color filters and in-camera processing.

There is metadata in the RAW about white balance, and possibly color profile. Outside of DPP, most RAW editors ignore any color profile metadata. Most will apply white balance, but again, the algorithms to achieve a given white balance in LR are different than in DPP which are different than in Aperture which are different than in RawThearapee.

Color is, if I had to distribute it among causes, is 95% math, 4% due to color cast due to color noise, and 1% hardware. We aren't talking about film here, where the chemical makeup of the film, or the chemical bath used to expose it, or the color characteristics of the lens matter. It's digital data. It's a matrix of numbers assigned color ranges like red, green, and blue. You can over-weight red, and under-weight green, and leave blue normally weighted, and completely change what the color looks like once the image is demosaiced and rendered.

It's all about the math.

 

[Marsu42]

The output in lightroom will be all about how LIGHTROOM processes the data. [...] Run the same experiment with DPP...you will get obviously different results compared to the Lightroom test

LR only has camera emulation modes (see "camera calibration, last item in Develop module) and doesn't claim to do 1:1 identical output to each and every camera - but looking at my 6d and 60d the results happen to be pretty close to each other.

As this camera calibration has such a vast impact on the look of the raw conversion, it's all about personal taste anyway unless you shoot with a color card to get a dng profile, "neutral" might con people into thinking it's "unprocessed" or "real".

 

[Valvebounce]

AAARRGGHH BUMP
That was me falling into that trap! ;D

Cheers, Graham.

"neutral" might con people into thinking it's "unprocessed" or "real".

 

[jrista]

The output in lightroom will be all about how LIGHTROOM processes the data. [...] Run the same experiment with DPP...you will get obviously different results compared to the Lightroom test

LR only has camera emulation modes (see "camera calibration, last item in Develop module) and doesn't claim to do 1:1 identical output to each and every camera - but looking at my 6d and 60d the results happen to be pretty close to each other.

As this camera calibration has such a vast impact on the look of the raw conversion, it's all about personal taste anyway unless you shoot with a color card to get a dng profile, "neutral" might con people into thinking it's "unprocessed" or "real".

Aye, LR/ACR are incapable of exactly reproducing the picture styles of the camera. That was entirely my point...that when you see "out of 'camera' color" in a photo you have loaded in Lightroom, your not seeing what the camera rendered...your seeing what Lightroom rendered instead. Similarly, even DPP will not produce exactly the same color as Canon's own cameras, as the curves are not 100% identical, nor is the precision of the hardware executing the algorithm identical.

What you get out of the camera as a JPEG is mostly the result of in-camera algorithms and tone curves that process the RAW data coming off the sensor. Compression algorithms affect the IQ, the color, the quality. Trying to exactly match the literal Out-Of-Camera JPEG with RAW would be difficult and time consuming.

Sure, there is a natural response of the silicon to light, and there are slight differences in the actual RAW ADU coming off the sensor due to differences in CFA. But the significance of those differences pales in comparison to the algorithms that convert those RAW pixel values into full color RGB values in an actual image rendered to a screen.

I am actually amazed at how neutral Samsung NX1 data seems to be when it's rendered (by anything, LR or Samsung's own RAW Converter software). It's the only camera I know of that renders blacks as deeply and purely without any visible color cast:

Even the Nikon D7200 and A7II both have some color cast, and the 6D has a lot of color cast. The NX1 doesn't have the lowest noise, but man, I would take that totally neutral color and super deep, rich blacks any day over all the rest. I don't know how they did that, but it's a pretty amazing feat if you ask me.