Thursday, April 29, 2010

LabPQR, what is it?

For several years, a group of researchers around Prof. Mitchell Rosen at RIT has been publishing papers on LabPQR. Now that he is publishing on different topics may be a good time to survey LabPQR.

While most color reproduction can be done with colorimetry, some problems require a spectral treatment. Examples include:

  • Spectral printer models
  • Mapping from one device to another
  • Fluorescent inks and/or media
  • Physical media models
  • Ink-media interactions
  • Security printing
  • More than 3 colorant hues (e.g., CMYKOGV)
  • Multiple illuminants (metamerism index minimization)
  • Mapping K generation between two different CMYK printers
  • Scanner and camera characterization

From an implementation point of view, processing spectral color is relatively simple, you just have n-dimensional arrays instead of triplets, plus a few linear algebra functions. The harder problem is when we would like to use an ICC type workflow also for spectral imaging, because the LUT used for the PCS will be huge:

Space needed to store a PCS LUT
bands in bands out levels per band
size [bytes]
30 K
145 M
700 G
8*10^27 G

Storing a multidimensional vector for each pixel is expensive. Can we project on a lower-dimensional vector space? Yes, because the spectra are relatively smooth. With principal component analysis (PCA), the dimension can be quite low: between 5 and 8. We have known how to deal with this for decades (e.g., Judd and McAdam for the Munsell chips in 1964), it just requires linearly more processing.

Rosen et al.'s proposal is to introduce in the ICC workflow such a lower-dimensional space, they call Interim Connection Space (ICS):

interim connection space

The next wish is to maintain compatibility with colorimetry. Can we deviate from the usual PCA method of choosing the largest eigenvectors and build on some other useful basis?

When defining the basis vectors for XYZ, the new basis was chosen so that one vector coincides with luminous efficiency V (λ), achieving compatibility of colorimetry with photometry. A 1995 proposal by Bernhard Hill et al. at RWTH Aachen was to incorporate three colorimetric dimensions, achieving compatibility of spectral technology with colorimetry:

color encoding proposed by Hill at RWTH Aachen

With this, Rosen et al. proposed the following approach they call LabPQR:

  1. Calculate an operator similar to matrix R using regression analysis on a specific printer (unconstrained), or matrix R directly (constrained)
  2. Calculate the residual using principal component analysis
  3. Calculate the tristimulus values XYZ
  4. Calculate PQR from the residual (3 largest EV)
  5. Calculate LabPQR from XYZPQR using the CIE equations for CIELAB

The convenient aspect of LabPQR is that it is very easy to visualize. Instead of a 6-dimensional space one can imagine the conventional CIELAB space, where each point in the color gamut, however, is actually a 3-dimensional gamut in PQR.

This is very similar to the way one does Fourier colorimetry with quaternions. A quaternion may be represented as a complex number with complex real and imaginary parts, the two complex numbers having different and orthogonal complex operators. This is known as the Cayley–Dickson form, after its 19th century discoverers.

Returning to LabPQR, the figure below shows how the encoding is verified. Assuming the requested color is in gamut, one first looks up the CIELAB point using the first 3 coordinated, receiving the PQR gamut at right for that CIELAB point. There one looks up the PQR coordinates for the residual basis functions. The lower part shows the reconstructed spectra.

verification of spectral color reproduction using LabPQR

For out-of-gamut colors the gamut mapping has of course to be done by tessellating the 6-dimensional LabPQR space, as there is no PQR gamut for an interpolated CIELAB value. However, the underlying math is a straightforward extension of interpolating tetrahedrons.

Rosen et al. propose two variants of LabPQR: constrained and unconstrained. In the constrained variant, the colorimetric transform is purely Matrix R, i.e., it is computed as A/(A'A) from the color matching functions A. In the unconstrained variant, the colorimetric transformation is calculated via least squares analysis over a number of tristimulus vectors for spectral values. In both variants, the PQR basis vectors are calculated from metameric black using PCA.

How do the two variants of LabPQR perform in terms of accuracy, i.e., what price in loss of accuracy do we pay for compatibility with conventional metamerism theory?

Based on simulations (no LUT), the constrained model is more accurate in general. However, for a single fixed printer, the unconstrained method allows the use of less principal components: LabPQ.

Seeing a Living Brain Seeing

"To find out more about the input signal, Konnerth and his colleagues observed a mouse in the act of seeing, with resolution that goes beyond a single nerve cell to a single synapse. They refined a method called two-photon fluorescence microscopy, which makes it possible to look up to half a millimeter into brain tissue and view not only an individual cell, but even its fine dendrites."

"And this," Konnerth says, "is where things get really exciting." The orientation neuron only sends output signals when, for example, the bar pattern moves from bottom to top. Evidently the neuron weighs the various input signals against each other and thus reduces the glut of incoming data to the most essential information needed for clear perception of motion."
Maybe I'm just getting old, but hasn't this been the basis of all connectionist models for the past 30 years?

Wednesday, April 28, 2010

Subtract, Add, Divide, Repeat. Eventually Print.

I'd forgotten that that the PRINT button shows up in the movie Pi.

Parallel Processing for Imaging Applications

IS&T/SPIE Electronic Imaging is the must attend event for all aspects of electronic imaging, including imaging systems, image processing, image quality, and algorithms. The Symposium Steering Committee, chaired by Sabine Süsstrunk, is starting a new exciting conference on Parallel Processing for Imaging Applications as part of the program track on Image Processing.

Sunday, April 18, 2010

Panchromatic Joan

The Passion of Joan of Arc was produced in 1928. It was made with panchromatic film. Some viewers have found themselves "enthralled in the drama of an 82-year-old creation". I was impressed with the film but was also struck by the tonal expressiveness of the close-ups of Joan.

A quick check of figure 13.4 of The manual of photography: photographic and digital imaging By R. E. Jacobson and it's clear that these monochrome flesh tones probably do differ from earlier black and white and subsequent color renderings.

Thursday, April 15, 2010

Wednesday, April 14, 2010

Or is it Aquamarine?

Dimitris (of, the delightful color naming print you sent arrived. It got here a while ago but this post is overdue - thanks for the print.

Testing Colours to Enhance Web Readability

According to the University of Washington’s Department of Ophthalmology, 2.8 million Americans are colour blind, which can express itself in many variations and degrees of severity. Colour perception problems are important considerations when developing web sites to ensure that all users have access to the content and the functionality of site.

Based on the contrast ratio algorithm, Level AA conformance requires text to have a contrast ratio of at least 4.5:1; larger text (18 point font or larger, or 14 point or larger if bold) a contrast ratio of at least 3:1. For Level AAA conformance, text requires a contrast ratio of at least 7:1; larger text a contrast ratio of at least 4.5:1. Incidental text or images of text that are not part of the user interface or are purely decorative and text that is part of a logo or brand are excluded from the colour contrast requirement

DIY conformance testing tools are available for Windoze and Mac OS X.


Monday, April 12, 2010

Color errors in printing

A couple of centuries ago, a gentleman had to diligently practice his fencing to survive. Today a technologist has to diligently practice his google-fu to survive.
Last week I noted in Vanity Publishing how a recent study had shown that for self-publishing and photo-books color accuracy is not an important quality factor. Does this mean we can just forget about color quality and invest our time into something more useful?

Vanity publications are only a small part of the total color printing market, and there are many customers that care a lot about color accuracy. In Silicon Valley's distorted reality field it is easy to get caught up in the hype and think everywhere it is like here. Sure, fresh MBAs turn their thesis on a printing related Web service into a start-up and then hire logistics experts to squeeze out good profits, or at least good investors, and think investing in good press people is a waste of resources, but in the real world there are many highly competent press people who make a decent living while contributing to society.

Therefore, I decided to practice my google-fu to learn what is out there. I could navigate to UGRA, FOGRA, GATF, TAGA, and all the other experts, but I was interested in learning what exists out there in the wild.

There are many competent color printers out there who know what they are doing and do not get confused by ICC profile versions. One who came out on Google's first result page was Ing. Rainer Wagner of WPC. His site is very competent and his data is consisted with my own experience. The fact that he openly shares it is another mark of his competence. So let us have a look at his data on tolerances in spectral color measurement.

tolerances in spectral measurement

The green line denotes the minima in his tolerances, while the maxima are in red. It is good engineering practice to first consider the big problems, then proceed towards the small potatoes (business people tend to go for the low-hanging fruit).

The biggest error source is the provenience of the ICC profiles. Most standards are not about how something should be done; rather, they specify how an output looks so the next entity in the workflow can process it. For example, JPEG does not specify how to encode an image, it specifies how to decode an image. Standard bodies enable to implement in an interoperable way the very best technology, but they never say anything of how the technology is implemented, other than giving examples to get interested parties started. The implementation is where businesses differentiate and compete.

In this spirit, the various providers of ICC profiles, as well as the producers of profiling software, have products at various quality levels so they can compete in every market. If they are interested in the high-end market, they will hire skilled color scientists; if they are just interested in making a buck in the mass-market, they may just hire cheap labor in a low-wage country.

Therefore, a good printer must be knowledgeable on the various sources of ICC profiles. When a job comes in, the printer will look at the embedded profiles to gauge their producer's reputation, then he can assess what quality level is realistic and manage the customer's expectations. If external profiles are specified, the good printer will have good profiles at hand.

Next down is the color difference between a proof printer and a press. This is a tough problem for offset and gravure printing, but in digital printing it is a non-issue, because the press is used for proofing.

I was surprised next on the list is the error introduced by color transformation algorithms in color management modules. I thought that after 25+ years all color transformation problems would have been solved. What happened? If you know, please leave a comment.

It is not a surprise there is no perfect agreement between instruments. Once at a lecture at RIT, Dr. Robert Hunt made the comparison that performing a color measurement is like placing a color chart in the fireplace, view it with a looking glass from the roof though the chimney, while the spouse would illuminate it by shining a flashlight from the garden.

Indeed, designing color measurement apparata is a very difficult art. What is the sphere diameter? Where are the baffles located? How big are they, etc. Measurements can only be compared if they are made with instruments using exactly the same head geometry.

Press drift is another number that surprises me. Online feedback loops based on viscosity measurements and online spectrophotometers should allow very tight tolerances. I am not a press person, so I cannot explain this large tolerance.

Today the drift of a proof printer is very small, and a diligent printer will not have much trouble keeping the drift below the perceptual threshold.

The tolerance for measurements with the same instrument can be made very small. However, it is surprising to see how often things get out of control. The instrument tilt is very critical, as is the diligent calibration. It is not only a matter of the instrument itself, but also of the operator. Therefore, techniques like collaborative testing and periodic absolute calibration are essential for high quality color printing (see here for details).

Color Imaging XVI: Call for Papers

The 2011 SPIE/IST Electronic Imaging Conference has posted calls for papers on the conference web site. As a memeber of the program commitee of the Color Imaging XVI: Displaying, Hardcopy, Processing, and Applications conference, please consider submitting a color imaging related abstract.

The conferences will be held 23 - 27 January 2011 at the Hyatt Regency San Francisco Airport Hotel in San Francisco, CA.

The dues date for abstract (500 words) and summaries (200 words) is 28 June 2010.

Wednesday, April 7, 2010

Vanity publishing

The latest full issue of the Journal of Electronic Imaging has a very interesting article by Jonathan Phillips, Peter Bajorski, Peter Burns, Erin Fredericks, and Mitchell Rosen: Comparing image quality of print-on-demand books and photobooks from web-based vendors.

Sunday, April 4, 2010

Bode's Law Doesn't Bode Well for Us

Both photometry and spectroscopy have been used to discover the more than 400 known exoplanets. The photometric light curve of a distant star with a transiting planet, when combined with detailed spectroscopic data, yields the planet radius, mass, and semimajor orbital distance, as well as the mass, radius, and effective temperature of the host star. The holy grail of this impressive scientific endeavor is, of course, to find Earth-like planets (possibly inhabited with Earth-like beings). However, the current odds of achieving such a find are not looking favorable because the trend indicates that Bode's law is being broken by the gaseous exo-giants.

Relative sizes of our planets (distances not to scale)
array of pyramidal quantum dots

Friday, April 2, 2010

World Wide Gamma

This is an experiment to estimate an average gamma or display non-linearity for the World Wide Web.

To participate, use the tool shown below to create an equal lightness step ramp. To create this ramp use the '+' and '-' buttons to lighten or darken the corresponding patches above the buttons. A black patch is shown furthest to the left, a white patch is shown furthest to the right and initially the intermediate patches are shown with random lightnesses.

When you are satisfied with your equal lightness step ramp press the 'plot' button to submit your data and view a plot of your results (black dots) versus the current average (red dots). You will also be able to submit optional feedback about your display and viewing conditions.

Thank you for your participation.

Note: The above background for the ramp consists of relatively coarse black and white squares in order to anchor the ramp to an approximate middle gray without using a solid gray, which would vary by display.

Jerusalem Color Stencil

"Time, we say is Lethe; but change of air is a similar draught, and if it works less thoroughly, does so more quickly." Der Zauberberg, T. Mann.

" a thing a name implies, if not passing judgment on it, at least defining it; that is to say, classifying it among the familiar and habitual..." Der Zauberberg, T. Mann.