scanin/scanin

Summary

Convert an 8 or 16 bit per component TIFF image of a test chart into .ti3 device values using automatic pattern recognition, or manual chart alignment.
Performs other tasks associated with turning a TIFF raster of test patches into numeric values.

Usage Summary

usage: scanin [options] input.tif recogin.cht valin.cie [diag.tif]
   :- inputs 'input.tif',  and outputs scanner 'input.ti3', or

usage: scanin -g [options] input.tif recogout.cht [diag.tif]
   :- outputs file 'recogout.cht', or

usage: scanin -o [options] input.tif recogin.cht [diag.tif]
   :- outputs file 'input.val', or

usage: scanin -c [options] input.tif recogin.cht scanprofile.[icm|mpp] pbase [diag.tif]
   :- inputs pbase.ti2 and outputs printer pbase.ti3, or

usage: scanin -r [options] input.tif recogin.cht pbase [diag.tif]
   :- inputs pbase.ti2+.ti3 and outputs pbase.ti3

 -g                   Generate a chart reference (.cht) file
 -o                   Output patch values in .val file
 -c                   Use image to measure color to convert printer pbase .ti2 to .ti3
 -ca                  Same as -c, but accumulates more values to pbase .ti3
                      from subsequent pages
 -r                   Replace device values in pbase .ti3
                      Default is to create a scanner .ti3 file
 -F x1,y1,x2,y2,x3,y3,x4,y4
                      Don't auto recognize, locate using four fiducual marks
 -p                   Compensate for perspective distortion
 -a                   Recognize chart in normal orientation only
                      Default is to recognize all possible chart angles
 -m                   Return true mean (default is robust mean)
 -G gamma             Approximate gamma encoding of image
 -v [n]               Verbosity level 0-9
 -d [ihvglLIcrsonap]   generate diagnostic output (try -dipn)
     i                 diag - B&W of input image
     h                 diag - Horizontal edge detection
     v                 diag - Vertical edge detection
     g                 diag - Groups detected
     l                 diag - Lines detected
     L                 diag - All lines detected
     I                 diag - lines used to improve fit
     c                 diag - lines perspective corrected
     r                 diag - lines rotated
     s                 diag - sample boxes rotated
     o                 diag - sample box outlines
     n                 diag - sample box names
     a                 diag - sample box areas
     p                 diag - pixel areas sampled
  -O outputfile       Override the default output filename & extension.

Usage Details and Discussion

scanin is setup to deal with a raster file that has been roughly cropped to a size that contains the test chart. It's exact orientation is not important [ie. there is usually no need to rotate or crop the image any more finely.] The reference files are normally set up with the assumption that the edges of the chart are visible within the image, and if the image is cropped to exclude the chart edges, it may well not recognize the chart properly. It is often better to crop out anything outside the chart itself (i.e. labeling text, logo's below the chart etc.) It is designed to cope with a variety of resolutions, and will cope with some degree of noise in the scan (due to screening artefacts on the original, or film grain), but it isn't really designed to accept very high resolution input. For anything over 1200 pixels on a side, you should consider down sampling the scan using a filtering down-sample, before submitting the file to scanin. Similarly, any file with a large level of noise (due to screening or scanner artefacts, or a noisy surrounding texture) should consider cropping out the noisy surrounding, or down sampling the image or filtering it with some average preserving filter before submitting it to scanin. Examining the diagnostic output (ie. -dig and -dil) may help in determining whether noise is an issue. To check that the chart has been correctly recognized, use -dipn and examine the diag image.

There are 5 basic modes that scanin operates in.
A number of flags and options are available, that are independent of the mode that scanin is in.

Normally scanin will try and recognize a chart, irrespective of its orientation. For charts that have some asymmetric patch size or arrangement (such as an IT8.7/2, or a chart generated by printtarg with the -s option), this is both flexible and reliable. Other charts may be symmetrical, and therefore having scanin figure out the orientation automatically is a problem if the recognition template does not contain expected patch values, since it will have an equal chance of orienting it incorrectly as correctly. To solve this, the -a flag can be used, and care taken to provide a raster file that is within 45 degrees of "no rotation".

Normally scanin will use automatic chart recognition to identify the location of the test patches and extract their values. If the chart CHT file  has four fiducial marks defined, then the chart can be manually aligned by specifying the pixel location of the four marks as arguments to the -F flag. The top left, top right, bottom right and bottom left fiducial marks X and Y co-ordinates should be specified as a single concatenated argument, separated by comma's, e.g: -F 10,20,435,22,432,239,10,239  The coodinates may be fractional using a decimal point. Four fiducial marks allows for compensation for perspective distortion.

By default the automatic chart recognition copes with rotation, scale and stretch in the chart image, making it suitable for charts that have been scanned, or shot squarely with a camera. If a chart has been shot not exactly facing the camera (perhaps to avoid reflection, or to get more even lighting), then it will suffer from perspective distortion as well. The -p flag enables automatic compensation for perspective distortion.

Normally scanin computes an average of the pixel values within a sample square, using a "robust" mean, that discards pixel values that are too far from the average ("outlier" pixel values). This is done in an attempt to discard value that are due to scanning artefacts such as dust, scratches etc. You can force scanin to return the true mean values for the sample squares that includes all the pixel values, by using the -m flag.

Normally scanin has reasonably robust feature recognition, but the default assumption is that the input chart has an approximately even visual distribution of patch values, and has been scanned and converted to a typical gamma 2.2 corrected image, meaning that the average patch pixel value is expected to be about 50%. If this is not the case (for instance if the input chart has been scanned with linear light or "raw" encoding), then it may enhance the image recognition to provide the approximate gamma encoding of the image. For instance, if linear light encoding ("Raw") is used, a -G value of 1.0 would be appropriate. Values less than 2.2 should be tried if the chart is particularly dark, or greater than 2.2 if the chart is particularly light. Generally it is only necessary to provide this is there are problems in recognizing the chart.

The -v flag enables extra verbosity in processing. This can aid debugging, if a chart fails to be recognized.

The -d flag enables the generation of an image recognition diagnostic raster. The name of diagnostic raster can be specified as the last in the command line, or if not, will default to diag.tif. Various flags control what is written to the diagnostic raster. Note that at least one flag must be specified for a diagnostic raster to be produced.
i    creates a black and white version of the input raster in the diagnostic output, to be able to compare with the feature extraction.
h    will show pixels in the input image classified as being on horizontal edges, in red.
v    will show pixels in the input image classified as being vertical edges, in green.
g    will show groups of pixels that will be used to estimate edge lines, each group in a different color.
l    will show valid lines estimated from the vertical and horizontal pixel groups, in white.
L    will show all lines (valid and invalid) estimated from the vertical and horizontal pixel groups, in white.
I    will show valid lines lines used to improve the final fit, in blue.
c    will show the lines with perspective correction applied in cyan.
r    will show the lines rotated to the reference chart orientation, in yellow.
s    will show the diagnostic sampling box edge outlines, rotated to the reference chart orientation, in orange.
o    will show all the sampling box edge outlines, in orange.
n    will show the ID names of the sampling boxes, plus the diagnostic sample boxes, using a simple stroke font, in orange.
a    will show the sampling areas as crossed boxes, plus the diagnostic sample boxes, in orange.
p    will show the sampling areas as colored pixels.

The combination of -dipn is usually a good place to start.

The TIFF file can be either 8 or 16 bits per color component, with 16 bit files being slower to process, but yielding more precise results.

If at all in doubt that the file has been recognized correctly, use the -dipn diagnostic flag combination, and check the resulting diagnostic raster file.
[ A badly recognised image will typically result in high self fit delta E's when used with colprof. ]

The -O parameter allows the output file name & extension to be specified independently of the last tiff filename. This works for the default, -g and -o modes. It is ignored for the -r, -c and -ca modes that use a basename for .ti2 in and .ti3 output. Note that the full filename must be specified, including the extension.