P45+ colour rendition

Thursday, 09 January 2014 19:59 administrator

Before buying my P45+ I was aware of some colour rendition issues. Sensors are different and may also give different colour rendition. It also used to be said that Capture One is needed to make the best of P45+ colour. 

Personally, I prefer my camera to give good colour without additional manipulation as I don't feel comfortable with manipulationg colour. So I investigated some of the options.

I also posted some early observations on this thread. The thread may have high noise to information ratio, but may be worth checking out.

The importance of white balance

Check the images below. Both were white balanced on the white paint on the left corner of the house, the left part was sunlit white and the right part was white in the shadow. White balance is absolutely crucial for colour rendition. It is a very good idea to shoot a WB target for each subject, but it is easy to forget.

The options

The major tools are Adobe DNG Profile Editor, Color Checker Passport and QPCard software. All those tools are free. QPCard has it's own reference card while the other two use XRite ColorChecker. The QPCard has more color samples and may work better in some cases.

Measuring colour rendition

There are standards for measuring colour. The adequacy of these standards may be up to discussion. Most standards are based on the Macbeth/Gretag/XRite ColorChecker. The standard ColorChecker has sixteen colour patches and six gray patches.

I shot the colour checker with my P45+ and also my Sony Alpha using a set of two small Elinchrome studio flashes and measured ΔΕ using Imatest. Processing was default, except that I adjusted exposure in raw processor to match reference data and adjusted saturation so Imatest reported 100%. 

Capture One

Camera ProfileDeltaESigma
P45+Capture One 7.458.4
P45+LR5.3, Adobe Standard 5.656.5
P45+LR5.3 DNG Profile Editor 4.715.31
P45+QPcard 6.977.85
Sony Alpha 99LR5.3, Adobe Standard 4.14.52
Sony Alpha 99LR5.3, DNGPE_2013_12_15 3.654.12

Based on these figures the Sony Alpha has much more accurate colours than the P45+. Also we can see LR 5.3 delivers more accurate colours than Capture One on the P45+. 

I am not sure how relevant these results are, however. In some of the cases the Capture One rendition is very good and the Adobe Standard is generally weak on the samples below. 

Imatest colour rendition charts

The two top left squares represent "dark" and "light" skin. 

Full size JPEG
Histogram: 20131224-CF044440_AS_colors.png
P45+ Adobe Standard
Full size JPEG
Histogram: 20131224-CF044440_C1_colors.png
P45+ Capture One
Full size JPEG
Histogram: 20131224-CF044440_DI_Adjusted_colors.png
P45 Adobe DNG Profile Ditor, DI, Adjusted colours
Full size JPEG
Histogram: 20131224-CF044440_P45Flash_colors.png
P45+ Adobe DNG Profile Editor (Studio Flash)
Full size JPEG
Histogram: 20131224-CF044440_QP_colors.png
P45+ QPCard profile
Full size JPEG
Histogram: 20131224-CF044440_colors.png
P45 Adobe Standard
Full size JPEG
Histogram: 20131225-_DSC3297_131225_colors.png
SLT99 Adobe DNG Profile Editor Dual Illuminant
Full size JPEG
Histogram: 20131225-_DSC3297_AS_colors.png
Adobe Standard

Imatest Color rendition diagram

Full size JPEG
Histogram: 20131224-CF044440_C1_colorerror.png
Capture One
Full size JPEG
Histogram: 20131224-CF044440_DI_Adjusted_colorerror.png
Adobe DNG Profile Dual Illuminant, adjusted
Full size JPEG
Histogram: 20131224-CF044440_P45Flash_colorerror.png
P45+ Adobe DNG Profile editor based on Studio flash
Full size JPEG
Histogram: 20131224-CF044440_QP_colorerror.png
P45+ QPCard
Full size JPEG
Histogram: 20131225-_DSC3297_131225_colorerror.png
SLT99 Adobe Standard
Full size JPEG
Histogram: 20131225-_DSC3297_AS_colorerror.png
Adobe Standard

Another test using an IT 8.7 target

I bought one of Wolf Faust's IT 8.7 targets and photographed it under studio flash lighting on both P45+ and Sony Alpha 99 SLT.

The merit of this test is that the reference target is different from the ColorChecker that calibration normally is made against, but also that many more fields are tested. The disadvantage is that the test target has a small colour space. A short description of the set up is given at the end of this page.

The raw images were developed using Capture One 7.2 and Lightroom 5.3. The table below shows the CIEDE 2000 DeltaE* values.

raw conversionP45+SLT99
Capture One FC5.175.4
Capture One linear3.474.54

Based on these tests I would say that there C1 with linear profile is slightly more accurate than LR5.3 for the P45+, but LR5.3 is significantly more accurate on the SLT 99. My guess is that the use of linear profile with C1 reduces the error in the L-coordinate.

DxO mark reports someting kalled SMI (Sensor Metamerism Index) this is measured on 8 or 16 patches of the ColourChecker under a given illumination and measures the capability of the sensor to produce a metameric match for those colours. It is calculated 100 - 5.5 Delta E* (mean). (needs to be checked). I calculated a "fake" SMI from my Delta E values.

The most accurate conversions were:

These two conversions are compared below, P45+ left SLT 99 right. Delta E* > 5 marked red

The image below shows the same conversion P45+ left and SLT 99 right, but with Delta E* > 3 marked in red

The last comparison is between P45+ converted with "film curve" and P45+ converted with my 131225 profile. In this case the LR 5.3 conversion is slightly better.

The impression I have from this test is that the Sony SLT 99 renders more accurate colours than the P45+, regardless the tool used. It also seems that the SLT 99 is considerable more accurate using LR5.3 conversion than using Capture One conversion.

The setup of the experiment

An IT 8.7 from Wolf faust was used. This was printed on matt Kodak paper. The target was accompanied by a CD containing a CGATS file describing the target. The data were mesaured for that batch of the taget but not individually measured.

The test target was set up alongside with a colour checker. Background was gery, and illumination was a Two Elinchrome D-Lite 4 flash flashes with soft boxes. Illumination was measured with a Minolta Flash Meter F and found to vary 0.1-0.2 EV max over the area.

A series of exposures were made and a an exposure selected that was well below clipping (1EV below). 

Raw processing was done so WB was set on the second brightest white patch on the colour checker taht also was used to adjust exposure. "Exposure" was adjusted so second brightest pathch had a Lab value of 0.80 in LR and RGB (200, 200, 200) in C1. In LR "Blacks" slider was adjusted to +25. 

Blacks were left as is in C1. When using "linear curve" in C1, all the gray patches had correct RGB values.

The two images were processed into 16 bits TIFF in Prophoto RGB. Patches were readout in CGATS format using Argyll CMS's "scanin -o " option. These files were loaded into Babelcolor's "Patchtool" and compared against CGATS data for the IT-8 target-

Innichen Church

White balance on all images is white wall a bit left of the left elbow of the sculpture. The rendition of the wall varies with colour profiles in LR. Although using same WB the QPCard profile renders varmer and the adobe profiles colder.

Full size JPEG
20130822-CF043946-4_QPFlash.jpg:No description found!
Full size JPEG
P45+ Adobe Standard
Full size JPEG
P45+ Capture One
Full size JPEG
P45+ Adobe DNG PE studio falsh
Full size JPEG
P45+ QPcard

Lago Antorno

Here I would suggest that the Adobe Standard profile (left top) renders grass as far to yellow. The white balance is same on all Lightroom images, the Capture One image (top right) has it's own light balance. These images were processed to taste.

Full size JPEG
P45+ Adobe Standard
Full size JPEG
P45+ Capture One
Full size JPEG
P45+ DNG Profile Editor (Studio Flash)
Full size JPEG
P45+ QPcard studio flash

Casual portrait

Full size JPEG
Adobe Standard for P45+
Full size JPEG
Capture One, defaults
Full size JPEG
This is converted by C1 with changes suggested by 'tho_mas'
on Lula, in this posting
Full size JPEG
Adobe DNG Editor
Calibrated with studio flash
Full size JPEG
QPCard with QPCard Software
Calibrated using studio flash


One of the conclusions I may draw is that the profiles may give different results under different conditions. In my view the Adobe Standard profile has often to much magenta and yellowish greens. QPCard may work best on subtle tones but also makes decent work on some landscape subjects. 

Although it seems that Capture One colours are the least accurate according to the Imatest measurement the results look good to me. 

I am quite happy with Lightroom. The option to genarate DNG Camera Profiles seems to be very useful, when default profile is not up to the task.

Last Updated on Sunday, 09 March 2014 21:24