Metadata for the preservation of colorimetry in digitization

0
532

Metadata are currently one of the pillars of digital preservation, understanding preservation as the ability to access and correctly understand a digital file at a future time. Metadata provide us with valuable intrinsic information about certain aspects of the data contained in our digital archive.

When we talk about digitization and finally about digital images or photographs, we have a series of information structures, such as the well-known EXIF, IPTC, or Adobe’s XMP. However, among these structures, there are very few references to the colorimetry area of our image.

Fortunately, the ICC color profiles embedded in our images present themselves as a good container of metadata to preserve valuable colorimetric information about the conditions under which the image was created alongside our digital image.

Within an input ICC color profile, there are usually fields such as ‘desc’ responsible for storing the device data for which the profile was generated, a ‘cprt’ or copyright field which can contain free text, and finally a field called manufacturer, described in the “Private and ICC Tag and CMM Registry (as of November 3, 2009)“, where it usually introduces the colorimetric information of our color chart used in the profiling process and the measurements resulting from that process.

colorLab In working with ArgyllCMS, this manufacturer-reserved field by default includes the colorimetric information in the ‘CIED’ tags associated with the manufacturer GretagMacbeth (currently X-RITE), the ‘DEVS’ tag associated with Monaco Systems (currently X-RITE), and the ICC’s own ‘TARG’ tag. The reason for containing this colorimetric information in these tags is that certain applications, such as the now-defunct ColorLab, can graphically reconstruct the color chart used, as well as obtain interesting colorimetric data for validating accuracy during profiling and deducing Delta Es.

The colorimetric information contained in these tags is described in a CGATS format, which in the case of ArgyllCMS is the complete TI3 file, resulting from operations with the Scanin tool; thus, we will frequently observe the following format:

BEGIN_DATA_FORMAT
 SAMPLE_ID XYZ_X XYZ_Y XYZ_Z RGB_R RGB_G RGB_B STDEV_R STDEV_G STDEV_B 
END_DATA_FORMAT

Where XYZ are the colorimetric values of our color chart in XYZ format, RGB the colorimetric values of our test image on the 100/255 scale, and the STDEV fields the standard deviation of the pixel reading results for each patch. The standard deviation in reading our pixel values is conditioned by noise, artifacts, or even dust residues on our image, which generate values far from the mean of each patch. For example, Scanin can work with two concepts: by default, a “robust” mean where only pixel values close to the mean are taken, and the true mean (option -m) which includes all values. High STDEV values probably indicate the presence of dirt, noise, or artifacts on a particular patch.

The Scanin tool can extract pixel values from an unscaled image, in the same format Photoshop would use via the -o option; thus, instead of a Ti3, we obtain a .val. However, we can also convert the RGB values of our Ti3 file through the following operations:

R = 35,451*(255/100) = 90.4
G = 29,001*(255/100) = 73.953
B = 25,601*(255/100) = 65.282

Although this strategy is carried out by various profiling tools, it is advisable to verify its implementation and, more importantly, ensure that our input profile is embedded within our image, since frequently, we have the habit of converting our images to working color spaces like Adobe RGB or Prophoto, causing this colorimetric information to be irretrievably lost.

The concept of whether or not to convert our images to working or display color spaces has sparked numerous discussions, as certain incorrect configurations in graphic editors can disregard these profiles, losing characterization information and thus severely altering the image. In this situation, we are addressing two distinct problems: on one hand, the file and on the other, image distribution. Our file must be clearly structured with clear preservation purposes; therefore, all the colorimetric information we can store together with our images will be of vital interest to face future reviews and interpretations of an image’s colorimetry.

A widespread habit in recent decades in the digitization of historical or heritage archives has been to accompany our images with a color scale such as Q13 or the IFRAO Scale in the field of archaeology or rock art as an attempt to record color. Unfortunately, by not accompanying our images or projects with the exact colorimetry of these color scales, it is impossible to perform a characterization of the scenes under which they were taken.