Evaluating Printer Profile Darkness Reproduction Capability

1881 Test Patches
These test patches are generated with PatchTool and only those patches within the range 7< L< 16 were chosen.

All 1881 patches are within the sRGB color spaces, but not all are within the printer gamut, as shown below.

Above show the test patches w.r.t the respective color spaces

Test Method

Result

Left: with MP              Center: BasICCOLOR                  Right: PM5

basICCOLOR DropRGB Profiler

An amazing Profiler, version 1.2 was released in 2006. Just drag&drop the measurement file and the icc profile can be generated. Following test was done with G7at which performs soft proofing of profile accuracy analytically.

Monaco_s9045-729_basicc.icm generated with dropRGB Profiler

Generated with Monaco Profiler, with the same measured data

RGB 8-bit vs 16-bit

How does bit depth affect the reproduction of the output image?

Below comparisons were performed with GRACoL2006_Coated1. 8-bit (per channel) and 16-bit (per channel) tiff files were output with PatchTool. RGB values were then extracted from the tiff files, read back and compared with the reference data.



sRGB



Adobe RGB

  

ProPhoto

Summary

Some color patches in GRACoL2006_Coated1 are out of sRGB color space, they are clipped and hence resulting in high dE error.





8-Bit Depth per Channel
            Average  90th %    95th %  
sRGB         1.27     0.72      9.47         
Adobe RGB    0.35     0.52      0.59  
ProPhoto RGB 0.44     0.73      0.88 

16-Bit Depth per Channel
            Average  90th %    95th %  
sRGB         1.04     0.38      9.45         
Adobe RGB    0.04     0.01      0.01
ProPhoto RGB 0.00     0.01      0.01

Retrieving a Monaco Session with CP2

Restore a Monaco session into CGATS format

Monaco session saves the colorimetric data in Lab values only, no spectral data was saved. Following procedure helps to recover the previously measured Lab values to CGATS file.

1. Open ColorPort2, go to Measure Target menu and select the appropriate target to be restored. Drag&drop the saved Monaco session data into CP2.

2. Save data in CGATS.
Do not save with PM5 option. The reason being that PM5 option saves both rgb and spectral data. The RGB data cannot be used for printer profiling.
CGATS option in CP2 saves data in RGB, Lab as well as spectral reflectance. Remember to select spectral value scale 0-100 as that is the default for PM5.
The spectral fields for both output options are null.

3. Open the saved file with a notepad editor. Delete the spectral field header, row count is optional, others untouched. Refer to the illustration for clarity.






4. Open PM5 ProfileMaker, assign the corresponding reference data and drag&drop the edited file. You are now ready to proceed to generate the icc profile.

Yet Another Way for Evaluating Profile Performance

The idea is to capture the image instead of printing it out and carrying out the physical measurement. With this 'soft' approach, neither ink nor paper is needed, and any profile can be evaluated without the presence of physical device. The RGB values for the patches will be extracted from the captured image and compared with the reference data.



Option for sRGB conversion



Errors (dE2000 @95th % max error)

dE1= 0.42 (error introduced by extraction of RGB values)

dE2= 0  (possible error due to raster capture utility)

dE3= 5.07 for PM5
dE3= 4.34 for MP

Evalute Profile Soft Proofing Capability

AtoB1 in the icc profile specifies, with colorimetric rendering intent, the required conversion for device data (RGB or CMKY) to PCS data (Lab, typically). The information related to the transformation is derived solely from the characteristic data gathered during the profiling process and the implementation is pretty unknown to the end user. The accuracy of the employed technique will greatly determine the profile soft proofing capability.

Methodology
The same set of measured data for creating of the printer icc profiles with PM5 and MP, was used. The 120-patch test charts were generated for each profile, measurements were compared with the reference data. 

1)  Gamutvision
By setting both the input (Profile 1 or 3) and output (profile 2 & 4) to the same printer profile, the accuracy of the conversion can be thus tested.
a)  Firstly, the test image was read according to the input file (the printer profile under test).  In effect, the PCS output at this stage is the PCS data for the printer.
b)  The input image is then mapped to to the output image according to the selected rendering intent. As the input profile is the same as output profile, no actual conversion should be taken place. Setting to either relative or absolute intents does not make any difference as the same conversion LUT table is used, BtoA1, I supposed.

In actual fact, above test is very similar to the round-trip test. (identical test?)


with built-in 24 checker image



with 120 test patches



with 1617 test patches

2)  ColorThink Pro 3:  Evaluate Profile Proofing


with CTP3, 120 test patches


3)  Hard Proofing
The best way to test a profile is to carry out a test print with the profile. It reflects the overall performance of the profile quality, and no way that we can test only the soft proofing portion of the printer profile.

References:
Evaluate Profile Proofing by ColorThink Pro 3
Using Gamutvision

Gamut Boundary: PM5 vs MP

Viewed with CTP3

Have you ever try to take a closer look of the device gamut boundary with ColorThink3? Wonder why the gamut boundary near L=0 for device profile generated with PM5 is so unstructured. With appropriate setting, the similar gamut boundary can be shown with Gamutvision.

Profile Graphing by ColorThink3  Myth 26
In CT3, gamut is plotted with AtoB transformation which is used normally for soft proofing, as oppose to some other graphers that employ "round-trip" transformation. Nevertheless, the calculated gamut volumes or overall gamut displayed do not deviate significantly for either methods used.

Refer to above plot, the center (vertex) of the white point is L=100 (not L=0, viewed from bottom to top). In general , the boundary vectors for MP is smooth and well distributed but that is not so for PM5.

All above illustrations have one thing in common, that they were all derived from the same set of measurement data.

Lab Gamut

http://www.brucelindbloom.com/index.html?Home.html

 Photos courtesy and copyright © Bruce Justin Lindbloom

Gamut volume comparison
                    Bruce         BabelColor
Apple RGB          798,403         894,338
sRGB               832,870         934,328
Adobe RGB (1998)   1,208,631       1,346,825
Lab                2,381,085       NA
ProPhoto           2,879,568       2,653,173

Some of the possible Lab values are illustrated as above. The white gamut boundary is the sRGB color space and it is shown for comparison purpose. All Lab values were generated with BabelColor PatchTool.

Selection of Test Patches: PM vs MP

Side by side comparison between the test patches for PM5 and MP is as shown above. The test patches for the profiling target for MP is evenly spread out in the RGB space whereas for PM5, more sample patches are centered along the gray axis (R=G=B). Following illustrates the test patches around the sRGB gamut boundary, they are well distributed for both profilers.

Monoca 343 Target for RGB Printer Profiling

3 built-in test charts are available for Monaco RGB printer profiling. The smallest test chart consists of 343 reference patches. These test charts are preassigned and its RGB reference set does not make known publicly. But one can still extract such information after the measured result is saved in Monaco session file in that the RGB data is stored in 16-bit RGB values. If the RGB reference data is altered by more than 1 count the profiler will reject the saved session data set.

The RGB reference points are equally spaced. For Monaco 343 test chart, each RGB channel was divided into 7 points from value 0 to 255, hence 7x7x7 or 343 test points.

The reference RGB, measured Lab data and the printer gamut which was derived from the measured Lab, are illustrated below.

Round Trip Test

Gamutvision Round Trip Test offers an effective way to evaluate the implementation consistency of the generated icc profile.

Following computations will be performed when actual RT test is executed.

Step 1: Map a printer profile's gamut boundary to L*a*b* using AToBn and that
            forms the Reference Lab values.
Step 2: Map the Input Lab values back to printer RGB (or CMYK) using BToAn.
Step 3: Map the printer RGB (or CMYK) values to L*a*b*, again with the AToBn
            and that constitutes the Target values.
Step 4: Compare Reference and Target Lab values.
            If the profile is performing flawlessly the result should be reversible and
            there will be little difference between them.



RT Test1:  3D/2D HL Color Difference Test (dE*ab, saturation=1)

  


RT Test2:  Image Analysis with 120-Test-Patch (dE2000)

 

Measurement Considerations

Minimum Patch Size

Implementation between spectrometers varies and depending on actual usage, the minimum patch size may be different. For instance, minimum patch size for i1Pro with patch recognition is 10x10 mm. The same i1Pro when mounted on an i1iO automated scanning table patch size can be reduced to 6x7 mm. Should users try to use an i1Pro to measure patch sizes smaller than the required 10x10 mm, they are bound to experience  inconsistent or incorrect measurements.

Distance between instrument and measuring surface
Distance between the measurement head of the instrument and the paper surface is crucial. Operate away from the optimal distance can lead to huge color deviations in measurement results. For i1Pro, the gap which controls by the manual positioning gadget or the scanning ruler, of approximately 40 to 50 um, serves as the allowable maximum distance. Third party rulers sold for the i1Pro can also be a source of errors, as many of them are not within the specification required for obtaining acceptable measurement results.

icc Profile for Printer

The tags, namely, B2A and A2B, represent the look-up-tables (LUTs) that the color conversion routine uses to translate the source color numbers to Lab numbers, and then from Lab numbers to the destination color numbers, respectively. Both source and destination color numbers are device dependent in that they are specific to that device. The Lab values, also known as Profile Connection Space, is device independent.

Source Profile --> A2B --> (Rendering Intent) --> Lab
Lab -->  B2A --> (Rendering Intent)--> Destination Profile

a) From Device to PCS (Profile Connection Space)
AToB0: transformation to achieve perceptual rendering.
AToB1: transformation to achieve colorimetric rendering.
AToB2: transformation to achieve saturation rendering.

a) From PCS to Device
BToA0: transformation to achieve perceptual rendering.
BToA1: transformation to achieve colorimetric rendering.
BToA2: transformation to achieve saturation rendering.

Ideal Color Space

It is practically impossible to find a perfect working color space.

Kodak ProPhoto RGB, Adobe RGB (1998) and sRGB are examples of gray balanced and perceptually uniform RGB working color spaces.

Gray balanced means that equal RGB numbers always produce a neutral gray whereas perceptual uniformity implies that the similar adjustment number produces approximately the same degree of visual change within the working color space.

Large color space may not be ideal, so does small color space
Large color space such as ProPhoto RGB contains huge gamut which cannot be fully visualised on today monitor screen. Any correction done with wide gamut cannot be checked until the actual result is printed out. Every output device too has its own gamut thus added uncertainty and complication to the overall color reproduction. On the other hand, if the working space is significantly smaller than the input device such as camera or scanner, the possible colors in that image would be clipped and permanently loss when converted to an output gamut.

An ideal space would be one that color difference is minimised when converting from one color space to the other or vice versa.

Printer's Gamut

Following are factors that determine the gamut for the printing devices:
a) Type of ink and paper formulation.
b) The characteristic of the printer
c) The repeatability of measurement tool.
d) The algorithm used to derive the icc profile.
e) The selection of patches used for building the profiling target.

Converting a Working Space

assigned to aRGB

sRGB

Assigning a Profile

. Neither RGB nor Lab values are altered.

. Image will look differently after new profile is assigned as RGB values are
  interpreted according to new assigned profile.

Converting a Working Space

If Rendering Intent is unchanged
. RGB values will change according to the
  new working space.
. Lab values will not change.

If Rendering Intent is changed
. RGB values will change according to the new working space and the new
  rendering intent used.
. Profile conversion compresses out-of-gamut colors, Lab values will be changed
  accordingly.