target/targen

Summary

Generate a profiling test target values .ti1 file.  targen is used to generate the device channel test point values for grayscale, RGB, CMY, CMYK or N-color output or display devices. 

Usage Summary

targen [options] outfile
 -v [level]      Verbose mode [optional verbose level, 1..n]
 -d col_comb     choose colorant combination from the following:
                     0: Print grey
                     1: Video grey
                     2: Print RGB
                     3: Video RGB
                     4: CMYK
                     5: CMY
                     6: CMYK + Light CM
                     7: CMYK + Light CMK
                     8: CMYK + Red + Blue
                     9: CMYK + Orange + Green
                    10: CMYK + Light CMK + Light Light K
                    11: CMYK + Orange + Green + Light CM
                    12: CMYK + Light CM + Medium CM
 -D colorant     Add or delete colorant from combination:
                     0: Additive
                     1: Cyan
                     2: Magenta
                     3: Yellow
                     4: Black
                     5: Orange
                     6: Red
                     7: Green
                     8: Blue
                     9: White
                     10: Light Cyan
                     11: Light Magenta
                     12: Light Yellow
                     13: Light Black
                     14: Medium Cyan
                     15: Medium Magenta
                     16: Medium Yellow
                     17: Medium Black
                     18: Light Light Black
 -G              Generate good optimzed points rather than Fast
 -e patches      White color test patches (default 4)
 -B patches      Black test patches (default 4 Grey/RGB, else 0)
  -s steps        Single channel steps (default 0)
 -g steps        Gray axis RGB or CMY steps (default 0)
 -m steps        Multidimensional device space cube steps (default 0)
 -b steps        Multidimensional body centered cubic steps (default 0)
  -f patches      Add iterative & adaptive full spread patches to total (default 836)
                     Default is Optimised Farthest Point Sampling (OFPS)
  -t                 Use incremental far point for full spread
  -r                 Use device space random for full spread
  -R                 Use perceptual space random for full spread
  -q                 Use device space-filling quasi-random for full spread
  -Q                 Use perceptual space-filling quasi-random for full spread
  -i                 Use device space body centered cubic grid for full spread
  -I                 Use perceptual space body centered cubic grid for full spread
  -a angle           Simplex grid angle 0.0 - 0.5 for B.C.C. grid, default 0.333300
  -A adaptation      Degree of adaptation of OFPS 0.0 - 1.0 (default 0.1, 1.0 if -c profile provided)
 -t              Use incremental far point for full spread (default iterative)
 -l ilimit       Total ink limit in %(default = none, or estimated from profile)
 -p power        Optional power-like value applied to all device values.
 -c profile      Optional device ICC or MPP pre-conditioning profile filename
 -N nemphasis    Degree of neutral axis patch concentration 0-1. (default 0.50)
 -V demphasis    dark region patch concentration 1.0-4.0 (default 1.0 = none)
  -F L,a,b,rad    Filter out samples outside Lab sphere.
 -w              Dump diagnostic outfile.wrl file (Lab locations)
 -W              Dump diagnostic outfile.wrl file (Device locations)
 outfile         Base name for output(.ti1)

Usage Details and Discussion

The number of target patches needs to be chosen, depending on the media size, the type of device, and the quality of profile required. For an inkjet device, something like 3000 test points or more is desirable for high quality profiles, while 500-1000 will probably suffice for a medium quality profile. A few hundred may be sufficient for a preliminary profile. Well behaved printing devices (such as a chemical proof, or a high quality printing press) may produce good profiles with 1000 to 2000 test points. Well behaved RGB devices such as CRT monitors may need only a few hundred points, if a shaper/matrix type profile is to be produced, while pseudo RGB printers, or other RGB devices that a CLUT type profile may be used with, should probably choose somewhere between 500 and 3000 patches. For 'N' color profile creation, 3000 or more test points should probably be used.

The -v flag turns on extra verbosity when generating patch values. Extra diagnostics and verbosity may be available if a parameter is provided with a value greater than 1.

The -d parameter sets the colorspace the test values will be generated in. Video gray space is assumed to be an additive space, where a zero device value will be black, and a maximum device value will be white. A print gray space is assumed to be a subtractive space, in which a zero device value will be white, and a maximum device value will be black. If no colorspace is specified, subtractive CMYK is assumed as a default.

The -D parameter modifies the colorspace set by -d by allowing individual colorants to be added or subtracted from the colorspace.

The -G flag changes the Incremental Far Point Distribution algorithm from fast to good mode. Fast mode uses a limited number of iterations to optimize the patch locations, while good mode strives for a more even patch distribution by using more iterations.

The composition of the test patches is controlled by the following flags and parameters:

The -e parameter sets the number of white colored test patches, defaulting to 4 if the -e flag isn't used. The white patches are usually very important in establishing white point that the ICC data is made relative to, so it improves robustness to use more than a single point.

The -B parameter sets the number of black colored test patches, defaulting to 4 if the -B flag isn't used and the colorspace is grey or RGB. The black point can be very important for characterizing additive color spaces, so measuring more than one black patch improves robustness over measuring just a single point.

The -s parameter sets the number of patches in a set of per colorant wedges. The steps are evenly spaced in device space by default, and the total number of test patches will be the number of colorants times the value specified with the -s flag. If the -p parameter is provided, then, then the steps will be distributed according to the power value. e.g. the option -s 5 will generate steps at 0.0 0.25 0.5 0.75 and 1.0, while the option -s 5 -p 2.0 will generate steps at 0.0 0.0625 0.25 0.5625 and 1.0. By default, no per colorant test wedge values are generated. When creating a test chart for a device that will be used as a source colorspace, it is often useful to generated some per colorant wedge values.

The -g parameter sets the number of patches in a set of combined (nominally gray) wedges. This will typically be equal RGB or CMY values, and by default will be equally spaced steps in device space. If the -p parameter is provided, then, then the steps will be distributed according to the power value. e.g. the option -g 5 will generate steps at 0.0 0.25 0.5 0.75 and 1.0, while the option -g 5 -p 2.0 will generate steps at 0.0 0.0625 0.25 0.5625 and 1.0. By default, no gray combination values are generated. When creating a test chart for a device that will be used as a source colorspace, it is often useful to generated some per colorant wedge values.

The -m parameter sets the edge size of the multidimensional grid of test values. The total number of patches of this type will be the -m parameter value to the power of the number of colorants. The grid steps are evenly spaced in device space by default, but if the -p parameter is provided, then, then the steps will be distributed according to the power value. e.g. the option -m 5 will generate steps at 0.0 0.25 0.5 0.75 and 1.0, while the option -m 5 -p 2.0 will generate steps at 0.0 0.0625 0.25 0.5625 and 1.0. By default, all the device primary color combinations that fall within the ink limit are generated..

The -b parameter sets the outer edge size of the multidimensional body centered grid of test values. The total number of patches of this type will be the -b parameter value to the power of the number of colorants plus the (number-1) to the power of the number of colorants. The grid steps are evenly spaced in device space by default, but if the -p parameter is provided, then, then the steps will be distributed according to the power value. A body centered grid is a regular grid (see -m) with another smaller regular grid within it, at the centers of the outer grid. This grid arrangement is more space efficient than a regular grid (ie. for a given number of test points, it fills the space better.)

The behavior of the -e, -s, -g -m and -b flags, is not to duplicate test values already created by a previous type.

The -f parameter sets the number of full spread test patches. Full spread patches are distributed according to the default or chosen algorithm. The default algorithm will optimize the point locations to minimize the distance from any point in device space, to the nearest sample point. This is called Optimized Farthest Point Sampling (OFPS) . This can be overridden by specifying the -t. -r, -R, -q, -i or -I flags. If the default OFPS algorithm is used, then adaptive test point distribution can be fully enabled by supplying a previous or typical profile with the -c option. The total number patches specified will include any patches generated using the -e, -s, -g -m and -b flags (i.e. full spread patches will be added to bring the total number of patches including those generated using the -e, -s, -g -m and -b flags up to the specified number). When there are more than four device channels, the full spread distribution algorithm can't deal with so many dimensions, and targen falls back on an incremental far point distribution algorithm by default, that doesn't generate such evenly spread points. This behaviour can be forced using the -t flag. A table of useful total patch counts for different paper sizes is shown below. Note that it's occasionally the case that the OFPS algorithm will fail to complete, or make very slow progress if the -c profile is poor, non-smooth, or has unusual behaviour. In these cases a different algorithm should be chosen (ie. -Q or -I), or perhaps a smoother or lower resolution ("quality") previous profile may overcome the problem.

The -t flag overrides the default full spread test patch algorithm, and makes use of the Incremental Far Point Distribution algorithm, which incrementally searches for test points that are as far away as possible from any existing points. This is used as the default for dimensions higher than 4.

The -r flag overrides the default full spread test patch algorithm, and chooses test points with an even random distribution in device space.

The -R flag overrides the default full spread test patch algorithm, and chooses test points with an even random distribution in perceptual space.

The -q flag overrides the default full spread test patch algorithm, and chooses test points with a quasi-random, space filling distribution in device space.

The -Q flag overrides the default full spread test patch algorithm, and chooses test points with a quasi-random, space filling distribution in perceptual space.

The -i flag overrides the default full spread test patch algorithm, and chooses test points with body centered cubic distribution in device space.

The -I flag overrides the default full spread test patch algorithm, and chooses test points with body centered cubic distribution in perceptual space.

The -a angle parameter sets the overall angle that the body centered grid distribution has.

The -A adaptation parameter sets the degree of adaptation to the known device characteristics, used by the default full spread OFPS algorithm. A profile should be provided using the -c parameter if adaptation is set above a low level. By default the adaptation is 0.1 (low), and 1.0 (maximum) if -c profile is provided, but these defaults can be overridden using this option. For instance, if the -c profile doesn't represent the device behavior very well, a lower adaption than 1.0 might be appropriate.

The -l flag and parameter sets a total ink limit (Total Area Coverage or TAC), which is adhered to for all the generated points. It is generally good practice to set a test chart ink limit at least 10% higher than the ink limit that will be applied when making the resulting profile. In the case of device cube points, this can generate extra test values that lie at the ink limit boundary. For gray wedge values, any that exceed the ink limit are omitted. Full spread test values are all generated to lie within the ink limit. Although it doesn't make much sense, this parameter has an affect on additive device spaces (such as RGB), but should not normally be used with such devices. The total ink limit value will be written to the .ti1 file, and carried through automatically to the .ti3 file, so that it can be used during profile creation. If a profile is provided using the -c flag, then this will be used to estimate an ink limit, if none is provided with the -l flag. Ink limits are, as far as possible, always in final calibrated device values, and the calibration curves from the profile provided to the -c flag will be used to estimate the equivalent limit in the underlying pre-calibration device space values that targen creates.

The -p flag and parameter sets a power-like value applied to all of the device values after they are generated, the spacer colors. This can be useful in creating calibration charts for very non-linearly behaved devices. A value greater than 1.0 will cause a tighter spacing of test values near device value 0.0, while a value less than 1.0 will cause a tighter spacing near device value 1.0. printcal will recommend a power-like value if the verbose option is used. [ Note that for Print RGB space this is reversed, since internally a Print RGB space is treated as a CMY space. ]. Note that the device model used to create the expected patch values will not take into account the applied power, nor will the more complex full spread algorithms correctly take into account the power in generating values up to the ink limits. (A power-like function is used, to avoid the excessive compression that a real power function would apply).

The -c flag and parameter is used to specify an ICC or MPP pre-conditioning profile, for estimating perceptual distances and colorspace curvature, used in optimizing the full spread test point placement,or in creating perceptualy spaced distributions. Normally a previous profile for this or a similar device will be used, or a simpler, preliminary profile will be created and used. If no such profile is specified, a default device space model is used. Note that this will only have an effect if an algorithm that uses perceptual placement (such as -R, -Q, -I or the default OFPS with an -A value > 0.0) is being used.

The -N nemphasis parameter allows changing the degree to which the patch distribution should emphasise the neutral axis. Since the neutral axis is regarded as the most visually critical are of the color space, it can help maximize the quality of the resulting profile to place more measurement patches in this region. This emphasis is only effective for perceptual patch distributions, and for the default OFPS distribution if the adaptation parameter is set to a high value. It is also most effective when a pre-conditioning profile is provided, since this is the only way that neutral can be determined. The default value of 0.5 provides an affect about twice the emphasis of the CIE94 Delta E formula.

The -V demphasis parameter allows changing the degree to which the patch distribution should emphasis dark region of the device response. Display devices used for video or film reproduction are typically viewed in dark viewing environments with no strong white reference, and typically employ a range of brightness levels in different scenes. This often means that the devices dark region response is of particular importance, so increasing the relative number of sample points in the dark region may improved the balance of accuracy of the resulting profile for video or film reproduction. This emphasis is only effective for perceptual patch distributions where a pre-conditioning profile is provided. The default value of 1.0 provides no emphasis of the dark regions. A value somewhere around 1.5 - 2.0 is a good place to start for video profile use. A scaled down version of the -V parameter will be passed on through the .ti3 file to colprof where it will set a default value for the corresponding colprof -V parameter.  Note that increasing the proportion of dark patches will typically lengthen the time that an instrument takes to read the whole chart. Emphasizing the dark region characterization will reduce the accuracy of measuring and modelling the lighter regions, given a fixed number of test points and profile quality/grid resolution. The parameter will also be used in an analogous way to the -p power value in changing the distribution of -s steps, -g steps, -m steps and -b steps patches.

The -F flag and parameters is used to define an L*a*b* sphere to filter the test points through. Only test points within the sphere (defined by it's center and radius) will be written to the .ti1 file. This can be good for targeting supplemental test points at a troublesome area of a device. The accuracy of the L*a*b* target will be best when the -c option is used to specify a reasonably accurate profile for the device. Note that the actual number of points generated can be hard to predict, and will depend on the type of generation used. All means of generating points except the -f N & -r, -R and -q will generate a smaller number of test points than expected. If the -f N & -r, -R and -q methods are used, then the target number of points will be achieved. For this reason, the -f N -q method is probably the easiest to use.

The -w flag causes a diagnostic VRML .wrl file to be created, in which the test points are plotted as small spheres in L*a*b* colorspace. Note that for a CMYK device, the point spacing may seem strange, since the extra K dimension is compressed into the 3 dimensional L*a*b* space. If the -W flag is given, the plot will be in device space, with only the first 3 dimensions of each point being plotted.

The final parameter on the command line is the base filename for the .ti1 output file. targen will add the .ti1 extension automatically.

Some typical total patch number/paper size combinations are shown below. These "magic" numbers are found by using printtarg to compute the row length and number of rows, and then adjusting the total number of patches to fill the last row or paper size, in a trial and error fashion.

     Size (mm/Standard Name),   No. Patches

DTP20:

      1 x A4        540
      2 x A4       1080
      3 x A4       1620
      4 x A4       2160

      1 x Letter    570
      2 x Letter   1140
      3 x Letter   1710
      4 x Letter   2280

DTP 22:

      1 x A4        782
      2 x A4       1564

      1 x Letter    736
      2 x Letter   1472

  DTP41:

      1 x A4                   375
      2 x A4                   750
      3 x A4                  1125
      4 x A4                  1500

      1 x Letter               345
      2 x Letter               690
      3 x Letter              1035
      4 x Letter              1380

      1 x A3                   836
      2 x A3                  1672

      1 x 11x17              780
      2 x 11x17             1560


  DTP51:

      1 x A4                   266
      2 x A4                   532
      3 x A4                   798
      4 x A4                  1064

      1 x Letter              252
      2 x Letter              504
      3 x Letter              756
      4 x Letter             1008

      1 x A3                  580
      2 x A3                 1160

      1 x 11x17             570
      2 x 11x17            1140

  SpectroScan with square patches:

      1 x A4R       1014
      2 x A4R       2028
      3 x A4R       3042
      4 x A4R       4056

      1 x LetterR    999
      2 x LetterR   1998
      3 x LetterR   2997
      4 x LetterR   3996

  SpectroScan with hexagonal patches:

      1 x A4R       1170
      2 x A4R       2340
      3 x A4R       3510
      4 x A4R       4680

      1 x LetterR   1092
      2 x LetterR   2184
      3 x LetterR   3276
      4 x LetterR   4368

  Eye-One Pro:

      1 x A4        441
      2 x A4        882
      3 x A4       1323
      4 x A4       1764

      1 x Letter    462
      2 x Letter    924
      3 x Letter   1386
      4 x Letter   1848

  ColorMunki:

      1 x A4          90
      2 x A4         180
      3 x A4         270
      4 x A4         360

      1 x Letter      98
      2 x Letter     196
      3 x Letter     294
      4 x Letter     392

  ColorMunki -h:

      1 x A4        210
      2 x A4        420
      3 x A4        630
      4 x A4        840

      1 x Letter    196
      2 x Letter    392
      3 x Letter    588
      4 x Letter    784

  Scanner (printtarg with -iSS -s options):

       1 x A4R      1014
       2 x A4R      2028
       3 x A4R      3042
       4 x A4R      4056

       1 x LetterR   962
       2 x LetterR  1924
       3 x LetterR  2886
       4 x LetterR  3848