Rob Trek Forum

To try to undestand the variance in color rendering, I just made a small test case between my m43 and my E-300, using auto white balance (the setting i normally use outside) with an X-Rite Colorchecker Passport.

The setting use indirect sun light in a partially obscured room, so to minimize as possible the lighting variation.
I took three picture in succession, with the em5 II, the em1 II then the E-300. I used the 25mm f 1.8 for the m43, the 14-54mm at 25mm f3.0 with the e300.

Here are the jpeg results using the cameras default settings

the em5 mk II the em1 mk II the e-300 With my eyes, the e-300 looks more close to my vision than the m43. Both m43 are very similar.

Addition: i look at these images using a factory calibrated BENQ PD2700U screen in HDR10 mode thru Windows 10 and a NVidia RTX graphic card. I assume the restituted colors are not too far of the image recorded values.

This raise the next question: is this a sensor difference, a white balance difference or both?

What makes an image colour rendering?

In order to analyse the differences, let take a look at the different elements that act in the generation of the final image:

• The scene defines what needs to be rendered,

• The light impacts the scene colour appearance due to its colour definition,

• The lens defines the field of view (so the amount of colour and colour intensity that need to be rendered) and can introduce subtle colour variation due to the glass and glass coating used,

• The camera acquisition part is the combination of the ISO settings, the aperture used, the shutter speeds, all of those having no impact as long as these value do not introduce overexposure, and then the sensor itself which as the key role in converting the scene appearance into pixel values.

• The camera picture processing converts the sensor output into a JPEG file using all the other parameters setting of the camera at that time, possibly including other type of modifications such as HDR, high res, focus stacking…

• The JPEG file contains the rendered scene using all the current camera settings

• The raw file normally contains the untouched sensor’s interpretation of the scene. This also contains the parameters settings of the camera at the time the picture was taken as well as a small JPEG representation of the scene with the applied parameters for pre-viewing.

• The raw to JPEG applications allow to reprocess the sensor capture to obtain new JPEG files. There are multiple applications, some from the camera manufacturer which are able to reproduce exactly the in-camera process but also allow you to change the photo time settings, other from specialized company that will generate the output using differently interpreted or additional processing parameters.

In our case, the light and scene are constant, so have no impact in the output differences.

The lens are different, but the differences they can introduce are subtle in the original photo, and, for the rest of this exercise, I will use the same lens on the M43 as on the 43 body.

We are then left with the different sensors and the camera parameters.

As we are not going to play with over exposed images here, we also can ignore the aperture, shutter speed and ISO parameters (well, this last one may impact the colour rendering, but this would probably be the result of a deficient sensor).

So we are left with the other parameters. To help us analyse the output of the capture, we can use the RAW file definition: As those files contain the original sensor outputs, as well as the camera parameters, this can help us understand the differences between the physical sensor and the cameras' processing.

Then we can use the raw to JPEG application to change various camera parameter to see the impact in the final image rendering.

I will use CaptureOne as an independent raw processing application and Olympus Workspace for a camera compatible raw to jpeg application for the M43 camera (as we can see in the original post, there is no need to differentiate between the em5.2 and the em1.2 as the colour rendition is very similar). I will also use the Olympus Master 2 raw to jpeg application to manipulate the 43 camera images if required (still works under windows 10).

White balance, the most important parameter for colour rendering


Hopefully, I will not dig into the colour theory here…

The camera sensor captures the scene using three colours that more or less matches the colour sensory cells of our eyes. These are the Red, Green and Blue colour (RGB).

In order to revert the colour cast that could be induced by the light(s) illuminating the scene, the camera use a set of parameters called the white balance that help adjust the sensor output to render as white the white objects of the scene.

A camera normally have multiple means to define this white balance: you can let the camera guess it automatically (as our vision system partially does, excepted that we have a ‘memory’ hint as our brain sometime knows which elements of the scene are really white), based on a context (sunny, shadow, cloudy, using a flash …), or a single numeric value called the colour temperature express in Kelvin degrees.

For the automated part, there is two methods, one that requires us to take a picture of a white object, which, when done with the same scene lighting, allows to analyse the scene illuminating light colour, and the other which lets the camera analyse the sensor output of each photo taken to find the most intensive colour components of the image to determine what the white colour should had been in the same (white being the most luminous combination of all the scene colours).

Some cameras also have an independent colour sensor used to determine the white balance independently of the sensor output and, more important, independently of the field of view created by the lens.

Update 2 is down below.

I think you're on the right approach. Obviously the e-300 is warmer. The sensor and the processing in-camera will affect the output. What I would do is set a custom white balance for each camera by setting it via the "Capture WB" method against the grey side of the x-rite color card.
This way you have a baseline "white". Then take a picture of the color tiles and compare the RGB color values. Most likely you'll see higher "Red" values.

Practice time

One of the nicest thing of the pre-set automated white balance (the WB obtained by taking a photo of a white object) is that this is determined in a direct relation of the sensor output, and, using a raw output file, we can see how the RGB light components colour values are distributed even if after the compensation is applied (there is no easy way using commercial tools to see the raw colour values distribution).

For this we need to:

• Pre-set a white balance using a white chart (I will use the X-Rite ColorChecker Passport grey chart, however, mine being far more old that 2 years, this is only use as a comparative mean, not to formally calibrate the cameras’ outputs).

• Take a raw photo of the same white chart (to obtain the sensor output used to determine the white balance and the correlated white balance parameter)

• An application that can give us access to the sensor output histogram with distinct traces for the RGB values. Here I will use Capture One that allows to see the three colours’ plot superposed, making the analysis easier (there is a 4th graph with a grey color which is the overall luminance distribution)

The histogram is a graph in which the horizontal axis represents the intensity values (min on the left to max on the right) and the vertical axis represents the number of pixels with this value.

The histograms represents the rendered image, not the sensor data, therefore for a perfect capture of a grey image with a perfect white balance with a camera perfectly aligned and a perfect sensor, such histogram would consist of a vertical line at the center of the graph (all the pixels have the mid color intensity) as the camera tries to adjust the overal image intensity to be in the middle of the camera sensitivity range.

Let start with the EM1 mark II. This is a professional camera, so I would assume this is a rigorously adjusted system and we should receive a clean histogram from this camera, which is the case:

- The histogram plot and camera determined balance temperature is: - The Capture One determined colour balance temperature would be: - We can then use C1 to enforce a fixed white balance: In the first graph above, we can see that the EM1 mk II is well behave, and even if some reflection and dark area were included in the picture, the various color graph are well on top of each other. The camera has aligned the intensity of the most rendered pixels close to the center of the graph

Capture one does a similar distribution, but seems to align the middle of the average of the most exposed pixels close to the center of the graph.


The E-300 is a consumer camera, with a known orange cast output, so we should see some vastly different outputs:

- The histogram plot and camera determined balance temperature is: - The Capture One determined colour balance temperature would be: - The C1 enforced a fixed white balance: In the first graph above, we can see that the E-300 has a far more wild intensity distribution and there is a wild range of color intensity toward the center. (This is most probably cause by electronic noise I was probably a little confuse when I wrote that, noise may play a role, but not that much, more in the Update 3.) The camera has aligned the intensity of the most numberous intense pixels (close to the center of the graph Does not read what i meant, so removed), while promoting a little the red color.

Capture one does a similar distribution, but seems to align the middle of the average of the most exposed pixels close to the center of the graph.

Now, when we enforce a specific while rendering,we would expect to have similar (histogram final image) results, but this is not the case at all. This seems to make the EM1 mk II more greenish and the E-300 more reddish. (This is very strange and will need some investigation.
Not Strange at all, as i change the white balance in opposite directions, but strange in the way the output colors are so different for the same white balance setting, more in the Update 3). Here is a side by side rendering of this enforce rendering, the E300 is on the left.
More to come, be patient.

Trick of the day: if you have a ColorChecker Passport, open it to the color square page, reduce the ambient light, lighten it with your browser screen and compare what you see against the photos in the top post: which one seems to be the closest match? The EM1 mk II or the E-300?

To just keep thing short: the difference in the general histograms shape of the update 2 where mainly due to a light reflection on the grey target as well as a less well behave colour noise of the E-300.

Using a more controlled artificial light, we obtain the following colour distribution from Capture One with the em1 mk 2:
and, for the e-300:
we see again that the e-300 is promoting a little the red colour, and the em1 mk 2 is promoting a little the blue color. This give us some hint in how we can shift the em1 mk 2 colour balance toward the e-300 rendering.

On the latest OMD cameras (from at least since the em5 mk 2), they are adjustments A+- and G+- on the side the white balance setting in the super control panel.

The setting A+- allows to customize the Blue-Yellow determination of the white balance (similar to a kelvin change), the G+- customizes the Green-Magenta settings, like the tint settings of capture one.

To make the em1 mk 2 match the e-300 we then need to change the A+- in the positive direction, and the G+- in the negative direction. We still need to determine the values. If you have JPEG files to compare, you can use the Olympus Workspace application: it allows jpeg white balance setting with two controls having the same settings capabilities (excepted that the G+- are inverted) found on the camera:
with a little try and error, i found that setting A to +3 and G to -3 was providing a better match of the E-300.

The EM1 mk 2 jpeg output with A+3 G-3: The E-300 jpeg output: The EM1 mk 2 jpeg output with A+-0 G+-0:
As a side note:

Rob suggested to set both camera to the same white balance. This works in making the two camera having the similar rendering, but not in making the EM1 mk 2 behave as the E-300 default auto white balance:


The EM1 mk 2 rendering at 5000K: the E-300 rendering at 5000K:

Just went outside for a walk with the E300 and the EM1mk2 and did some comparative shooting. The E300 are placed first (all in P mode).

I think if you took the side by side comparison of the grey card and showed them individually, they would look more grey to the eye.
The e-300 seems to be pushing about 10% more red, while the green and blue channels are pretty equal. But is the "enforced" wb rendering by C1?
I compared the rgb values from each grey card in your image and found 25 bits difference (out of 255) between the two cameras.

I'm trying to understand the terminology you're using. It looks to be consistent, but I need to have a clearer definition.

"Camera determined" = Auto WB?
"C1 determined" = C1 Auto WB?
"C1 enforced" = C1 eye dropper?

Thanks!

Rob, your are spot on for the terminology i used.

I will add that for the camera auto white balance, this is really setting a custom white balance using the gray plate of the ColorChecker, then taking a photo of the same plate with this setting. As the white balance setting is a process taking account over all the pixel captured by the sensor, the photo taken with this white balance shows the tendancies resulting of the process (either due to the hardware or the software parts), which are different between both cameras. This is what we see then when looking to the histogram (i hope this is somehow understandable) .

Excepted for the part where i enforced the white color balance with C1, the jpeg are straight out of the camera jpeg, no c1 processing at all.

The real exercise was to find a way to 'see' the difference between the camera auto white balance and derive an idea how to change this so the output of the E-300 could be obtain using the EM1mk2 sooc.

This was pure curiosity from my side and totally unprepared exercise, i almost did it real time

Since, i also looked into the warm color preservation when using the auto white balance, it helps a bit.