DE102013011954A1 - Method and apparatus for camera-based characterization of the color rendering of displays - Google Patents

Abstract

Method and apparatus are used to capture the generated outputs of displays (especially monitors and projectors) with respect to a device-independent color space such as CIE XYZ. A set of input color value output measurement pairs allow the characterization of the color rendering of a display. The present invention reduces the time required for the measurement and in practice allows the use of significantly higher sample numbers. For measurement, input color values (samples) are combined into two-dimensional patterns. All samples of a pattern are simultaneously displayed on the display and recorded simultaneously, with multiple subpatterns being generated for each pattern: two reference subpatterns each contain only repetitions of a sample, several color subpatterns contain the samples of the pattern in a pseudorandom permutation, supplemented by a few repetitions of the samples of the associated reference pattern. Based on the reference samples of the color subpatterns, the measured values of the reference subpatterns are related to those of the color subpatterns, and the measured values for the color samples relative to the reference samples are expressed and reinterpreted with respect to another measurement position, so that the position- and angle-dependent variations of the color and brightness reproduction of the display are compensated and the equivalent of a traditional spot measurement arises.

Description

Technical area

The invention relates to a method for measuring the color outputs generated by a display as a function of the input color values and to a device for carrying out this measurement. The invention is for the color calibration of display systems such as LCD monitors, TVs and projectors and segmented display systems, which are composed of several individual displays.

State of the art

Color calibration of displays is essential for many applications: Professional image and video processing requires output devices that can reproduce the color space used for processing as accurately as possible and without distortion. In addition to this absolute color calibration is desirable for segmented devices that consist of a large number of individual displays - such as arrays of LCD displays or multi-segment projection systems - a color matching for the largest possible common gamut of all involved individual displays and each display on this common Gamut must be calibrated, even if it is not necessarily a calibration with respect to a precisely predetermined color space must be made.

• • Spektralphotometer, die die spektrale Energieverteilung der abgegebenen elektromagnetischen Strahlung im sichtbaren Bereich erfassen und die genauesten Messergebnisse liefern.
• • Tristimulus-Kolorimeter, die lediglich über drei verschiedene Farbfilter verfügen und einfallende Strahlung als Tristimulus-Wert bezüglich der durch die Farbfilter festgelegten spektralen Intensitätskurve erfassen. Dieses Wirkprinzip entspricht der Funktionsweise des menschlichen Auges. Mit Verwendung hochwertiger Farbfilter lassen sich direkt die XYZ-Werte erfassen.
• • Verwendung von Digitalkameras. Eine solche Kamera arbeitet nach dem gleichen Prinzip wie ein Tristimulus-Kolorimeter. Jedoch sind Kameras nicht auf absolute Messgenauigkeit optimiert.

Systems and methods for color calibration have existed for several decades. The gamut of devices can be described by color profiles, such as the standards administered by the International Color Consortium (ICC) [1]. Existing color management systems such as Adobe Color Management Modules (CMM) or the Windows Color System (WCS) integrated into the Windows operating system allow image content to be converted between different devices if color profiles are available. The creation of the color profiles requires the measurement of the actual state, ie the acquisition of the outputs of the display when displaying certain input color values. This acquisition takes place in a device-independent color space such as CIE XYZ [2], from which the human color sensation can be derived and which allows the conversion of the color values between different devices as a common reference color space. For measurement are used:

• • Spectrophotometers, which record the spectral energy distribution of the emitted electromagnetic radiation in the visible range and provide the most accurate measurement results.
• • Tristimulus colorimeters, which have only three different color filters and detect incident radiation as a tristimulus value with respect to the spectral intensity curve defined by the color filters. This principle of action corresponds to the functioning of the human eye. Using high-quality color filters, the XYZ values can be directly recorded.
• • Using digital cameras. Such a camera works on the same principle as a tristimulus colorimeter. However, cameras are not optimized for absolute measurement accuracy.

The measurement of a display output takes about 3 s to 60 s, depending on the measuring device used and the accuracy requirements. For this reason, the number of color samples to be detected is reduced to a minimum. Classic RGB displays use additive color mixing of the three primary colors red, green, and blue to produce the output. These displays are controlled by RGB signals.

In the following, the term color transfer function shall be used as the map F (r, g, b) of input values as RGB tristimulus on the generated output color values (X, Y, Z). The color calibration requires the determination of F and the inverse mapping F ^-1 .

• • M: 3×3 Matrix, die die XYZ-Koordinaten der Primärfarben für RGB beschreibt
• • k: dreidimensionaler Vektor, der den Schwarzpunkt in XYZ-Koordinaten beschreibt
• • drei separate Intensitäts-Transferfunktionen T_i(x), die die Abbildung zwischen Eingabe-Intensität des Kanals und Ausgabe-Intensität (normiert auf Intervall [0, 1]) beschreiben. T stellt meist keine lineare Funktion dar, sondern meist eine Potenzfunktion der Art T_i(x) = x^γ, wobei γ eine Gammakorrektur beschreibt, die der nichtlinearen Charakteristiken der menschlichen Helligkeitsempfindung nachgebildet ist.

Ideally, it can be assumed that the channels are independent, so that the red input channel only affects the intensity of the red primary color used for the output, analogous to green and blue (see also [3]). Under these conditions, the color transfer function can be decomposed into the proportions: F (r, g, b) = M * (T _r (r), T _g (g), T _b (b)) ^T + k

• • M: 3 × 3 matrix that describes the XYZ coordinates of the primary colors for RGB
• • k: three-dimensional vector that describes the black point in XYZ coordinates
• • three separate intensity transfer functions T _i (x) describing the mapping between input intensity of the channel and output intensity (normalized to interval [0, 1]). T is usually not a linear function, but usually a power function of the type T _i (x) = x ^γ , where γ describes a gamma correction that is modeled on the non-linear characteristics of the human brightness sensation.

The acquisition of M requires 3 individual measurements, k another one. If the display perfectly converts a gamma value, theoretically one more measurement per color channel is sufficient to completely determine T _i . In practice, most displays deviate from such an ideal image, so that n intensity levels must be measured per color channel. In extreme cases, all intensity levels For displays with 8-bit color depth, this requires 3 x 256 = 768 measurements.

Description of the problem / disadvantages of previous methods

The independence of the input color channels is no longer mandatory for many modern display technologies. Additional basic colors such as yellow, white or cyan are often used. In particular, projectors with DLP technology often work with color wheels on which in addition to the three RGB segments, several other segments are available. Some LCD TVs and projectors also use an additional base color to produce yellows. In the following, m> 3 denotes the number of primary colors.

While device-internal decomposition of the color transfer function in m intensity transfer functions would be possible, this no longer applies to the input signals, which are still transmitted as RGB data. The conversion to the intensities of the internally used color channels is done internally and can not be assumed to be known. For the measurement, this means that neither the exact XYZ coordinates of all m primary colors nor the internal intensity transfer functions can be determined. The representation of yellow can z. B. can be realized as a mixture of red and green and the additional yellow, the mixing ratios are neither known, nor can be calculated from the XYZ tristimulus values, since a three-dimensional space can not be clearly described with respect to a base with more than three dimensions ,

For measurement of F, it is no longer possible to resort to 3 × n or m × n individual measurements, but n ³ measurements result. In practice, therefore, the number of intensity levels to be measured is severely limited; typical values are between 8 (512 samples) and 17 (4913 samples). However, the use of n> 12 already leads to a very high measurement duration.

In [4] it is proposed for DLP projectors with additional white segmentation to first use the standard procedure assuming channel independence and additionally to use a sparsely populated three-dimensional look-up table that describes the deviations from this model and allows their correction , with a sampling density of 5 ^{3 is} recommended.

Sajadi et al. proposed the ADICT method in 2010, in which the function F or its inverse are modeled as multidimensional Bézier functions [5]. With this model, you could already achieve comparable results for n = 9 for linear interpolation of measurements with n = 16 for many devices.

However, the existing methods are not suitable for all types of displays. In particular, DLP projectors with more than 4 segments in the color wheel often operate with discrete thresholds beyond which additional segments are used. This often results in very local distortions or even discontinuities that can not or not sufficiently be detected with a low sampling density. The increase of the sampling density is only a theoretical option, since the time expenditure grows cubically with the parameter n. Sajadi and Majumder give for n = 9 about 2.5 hours and 11 hours for n = 16 as the measurement period, the latter regard as no longer practicable.

goal

• • Reduktion des Zeitbedarfs zur Farbkalibrierung von Displaysystemen, deren Farb-Transferfunktion sich mit einer vergleichsweise geringen Sampledichte (z. B. 17³) erfassen lassen.
• • Ermöglichung der Messung mit wesentlich erhöhter Sampledichte (z. B. 64³) innerhalb einer vertretbaren Messdauer (z. B. 5 Stunden). Bisherige Verfahren würden für eine derartige hohe Sampleanzahl (262144) mehrere Wochen benötigen.

The aim of the invention is to significantly accelerate the detection of the three-dimensional color transfer function. This allows two improvements over the prior art:

• • Reduction of the time required for color calibration of display systems whose color transfer function can be detected with a comparatively low sample density (eg 17 ³ ).
• • Allowing for measurement with significantly increased sample density (eg 64 ³ ) within a reasonable measurement period (eg 5 hours). Previous methods would require several weeks for such a high number of samples (262,144).

Description of the invention

construction

• • einem, zwei oder drei Rechnersystemen, wobei ein Rechnersystem mit mindestens einem zur Ansteuerung des zu messenden Displays geeigneten Videoausgang (HDMI/DVI/DisplayPort/VGA) ausgestattet ist
• • einer digitalen Spiegelreflexkamera
• • optional einem Kolorimeter oder Spektralphotometer
• • bei Verwendung von mehreren Rechnersystemen: einem Computernetzwerk zur Verbindung aller beteiligen Systeme
• • einer Softwarekomponente zur Ansteuerung des Displays mit Eingabebildpattern
• • einer Softwarekomponente zur Ansteuerung der Digitalkamera und zum Auslesen der Bilddaten
• • einer Softwarekomponente zur Durchführung der Messung.

The measuring device consists of

• • one, two or three computer systems, whereby a computer system is equipped with at least one video output (HDMI / DVI / DisplayPort / VGA) suitable for controlling the display to be measured
• • a digital SLR camera
• • optionally a colorimeter or spectrophotometer
• • using multiple computer systems: a computer network connecting all the systems involved
• • a software component for controlling the display with input image patterns
• • a software component for controlling the digital camera and reading the image data
• • a software component to perform the measurement.

preparation

For the measurement, the camera must be positioned in such a way that the image area of the display can be recorded completely and over the largest possible area. First, a geometric calibration is performed to determine the mapping between pixel coordinates of the camera and the display. Such methods already exist and are not part of the invention. It should be noted, however, that in the case of camera-based geometric calibration it is usually desired to decompose distortions caused by the display and the camera, while only the overall distortion is relevant for this application, and thus some partial steps of the geometric calibration procedure may be omitted.

Division into patterns and subpatterns

Let S be the set of all color samples to be captured in the RGB space. It is assumed that the samples are arranged in an (approximately) regular grid in the RGB space. The set of all possible color samples is represented by a cube in the RGB space. This cube is divided into s (approximately) equal slices along each dimension. There are s ³ partial cubes. All samples of S that lie within such a partial cube are combined into a pattern.

This division is based on the assumption that even if the display reflects the RGB space only distorted, still preserves the basic characteristics of the RGB model, otherwise an RGB input image would lead to a completely unsatisfactory display output. Taking this fact into account, the splitting results in samples of similar color and intensity being combined within a pattern, limiting the dynamic range within the pattern. This is especially true for the dynamic range between different color channels.

Each pattern produces a number of samples. Based on the above assumptions, this number varies only slightly between patterns. For the measurement, the parameters p and q are defined, so that the following applies: ∀i: (p - 1) · (q - 1) ≥ #Samples (i)

Let #Samples (i) be the number of samples in Pattern i Under this constraint the difference p - q is minimized.

For each pattern, a minimum and maximum RGB value is first determined by determining the minimum and maximum intensity of the occurring samples independently for each color channel. For each pattern several subpatterns are created. Each subpattern consists of a two-dimensional array of (p + 2) * (q + 2) color samples. The first subpattern consists exclusively of repetitions of the minimum sample, the second subpattern consists exclusively of repetitions of the maximum sample. Each additional subpattern consists of (p - 1) · (q - 1) color samples. If the pattern consists of fewer samples, the rest of the fields are padded with repetitions of the samples. The remaining fields for the number p · q are filled in half with repetitions of the minimum sample and half with repetitions of the maximum sample. The resulting samples are permuted pseudo-randomly, whereby the outer edge is not occupied, but only after the permutation with copies of each next inner color samples is filled. The first two subpatterns of a pattern are named their reference subpatterns, the remaining color subpatterns.

Camera-based measurement

For the measurement, a subpattern is shown on the display, whereby the p · q color samples are each displayed as rectangles colored with the respective sample value. The size of the rectangles results from the resolution of the display and the parameters p and q and is chosen so that the available image area is utilized as completely as possible.

The measurement itself is done by means of the digital camera. Most digital cameras use color filters with color filters placed in front of the actual sensor elements. In this case, a sensor element can detect only a part of the tristimulus value which is arranged by the respective color filter. Therefore, locally all color filters are repeated in a fixed order. Usually a so-called Bayer pattern with RGB filters is used, whereby in each 2 × 2 block on one diagonal there are two green filters each and in the remaining two sensor elements one red and one blue filter each.

For the measurement method, a lookup table is precalculated for geometric calibration, which assigns a color sample and a color channel to each raw camera pixel, and thus also takes into account the arrangement of the color filters. Furthermore, each camera pixel is assigned a weight value, with the weights always decreasing to the edge of the image rectangle of a color sample. This will prevent blurring, geometrical calibration errors or (caused for example by thermal effects) Changes during the measurement lead to situations in which neighboring color samples influence.

Each camera photo produces a measured value for each color sample and each color channel of the camera as a weighted average of all camera pixels that capture the rectangle of this sample displayed by the display and correspond to the respective camera color channel. The measurements used refer to the raw sensor data provided by the camera's CMOS or CCD chip and should have undergone as few post-processing steps as possible. The knives should be viewed relative to the exposure time, f-number, and ISO sensitivity used during shooting. Furthermore, both the vignetting effect of the camera lens and the uneven illumination of the display lead to position dependencies of the measured values, which must be compensated later.

• • Neben der Belichtungszeit wird auch die ISO-Empfindlichkeit variiert
• • Die relativen Faktoren der unterschiedlichen Belichtungszeiten und ISO-Werte zwischen zwei Fotos wird nicht als konstant angenommen, sondern aus der Überlappung von in beiden Aufnahmen gut ausbelichteten (also weder unter- noch überbelichteten) Farbsamples rekonstruiert.

Capturing a subpattern usually requires multiple photos. In order to cover the full dynamic range, a modified High Dynamic Range (HDR) method is used. The standard HDR process takes multiple shots of the same scene with different exposure times. From the knowledge of the relative differences in the exposure time, the individual images can be combined to a total recording with a higher dynamic range [6]. As part of the measurement procedure, two modifications were made to the standard procedure:

• • In addition to the exposure time, the ISO sensitivity is also varied
• • The relative factors of the different exposure times and ISO values between two photos are not assumed to be constant, but are reconstructed from the overlap of well-exposed (ie neither underexposed nor overexposed) color samples in both images.

Exposure control is automatic. For this purpose, a path is defined by the parameter space spanned by exposure time and ISO value. The measurement starts with initial parameters. If overexposed color samples are present, the next point along the path will be selected and another shot taken. This is repeated until no overexposure occurs. The procedure is analogous to underexposure, wherein the path is then traversed from the path point during the initial exposure in the other direction. Contrary to the requirements of ordinary photography, the exposure is done separately for each color channel of the camera. A very intense red color sample has a very low, but not negligible, value with regard to the blue color filter of the camera, so that, if necessary, long exposure times may be necessary for its detection.

To improve quality, a minimum number of repetitions is set. Any photo that contained correctly exposed samples in any color channel will then be repeated multiple times. For the combination in the HDR method, the data of photos are averaged with the same exposure parameters to further reduce the measurement noise. For each Subpattern thus several photos are made. After combining with the modified HDR method, p · q measurements are made per color channel of the camera, the outer edge is ignored during processing. The combination is iterative, in which the measurement data is accumulated in each case. Starting from the exposure level that contains the most correctly exposed color samples, an exposure factor is first calculated using the known parameter values for exposure time, aperture and ISO value, with which the data can be converted into an absolute color space:

Here A denotes the f-number, Δt the exposure time and I the ISO value. The measured values of the camera are multiplied by the factor I, the resulting values are proportional to the incident on the camera irradiance E, but weighted by the spectral sensitivity function of the color filter.

This conversion represents the initial values for the accumulation. Each additional exposure level is added by selecting the one whose number of overlapping samples, that is, samples correctly exposed in both the current accumulation and the photo, is maximum. Due to the overlapping sample values, a model function f (x) = αx + β is fitted so that the error in the imaging of the exposure level is minimized. α corresponds to the relative factor, which in the ideal case is called the ratio l₁ / l₂ should be the two exposure factors, but slightly different in practice. In addition, β permits a correction of residual currents which also flow with no input radiation. The accumulation of overlapping samples is performed with a weighted average, the weighting as in the standard HDR method depending on the intensity value of the camera. Values near the under- or overexposure receive a lower weight, values in the middle of the value range a higher weight. The HDR process is performed separately for each color channel of the camera.

To measure a pattern, all individual measurements of the subpatterns are combined. The two reference subpatterns each contain only one reference color sample and compensate for the positional dependency of the measured values and the consideration of the background light during the measurement. It should be noted in particular that the display output itself is reflected diffusely in the room and partly reaches back to the display. This effect results in a dynamic background light caused by the samples within the subpattern itself.

The color subpatterns contain a set of repetitions of the minimum and maximum reference values of the pattern distributed in a pseudo-random manner. The method works with a linear model of the form f (x) = αx + β. Due to the measurements for the reference samples in the color subpattern and the measurements for the same reference samples in the respective reference subpattern at the same position, the parameters α and β can be separated for each color channel and each of the two reference subpatters a minimization procedure will be correct. It is thus possible to convert the measured values of the reference subpattern so that the expected measured values for the reference samples can be estimated at the dynamic background light of the color subpattern at each point of the subpattern. In a sense, the reference subpatterns are linearly transformed into the measured values of the color subpatterns, so that the error to the reference samples acquired in the color subpattern is minimized and the brightness characteristic for the complete display is approximated.

The measured values are now converted relative to the reference values. Let x _{i be} the camera reading for a sample at position i and a camera color channel. Then with R _{i, min} and R _{i, max} an estimate for the minimum and maximum reference value at position i is available for this color channel and x _i can be expressed relative to these values as:

Furthermore, x _i can be reinterpreted to a measured value x ' _i which the camera would have detected if the sample had not been measured at position i but at position j: x ' _i = r _i * (R _{j, max} - R _{j, min} )

In this way, the equivalent of a traditional spot measurement is created as with a spectrophotometer or colorimeter. As a reference point j, a sample near the center of the image area is used.

The resulting x ' _r readings can be directly assigned to the measured input sample. If more than one color subpattern is available, all results of a sample are averaged. In this way, the measurement noise is additionally reduced, the position-dependent measurement differences further reduced and also compensates for disturbing influences of adjacent color samples. The number of color subpatterns per pattern allows you to balance the speed and quality of the measurement.

In addition to the s ³ patterns, we continue to insert an additional reference pattern. This only contains the minimum and maximum samples of all other patterns repeated several times. The decomposition of this pattern into subpatterns as well as the measurement itself is done exactly as described above. However, this pattern will be measured first. In this way, measurement data for the reference values are generated in a common recording and with a constant background light. The measurement results for the reference samples are denoted in each case as R _{ref, min} and R '. The processing of the subpatterns for all ordinary patterns then takes place with respect to these reference values x '' _i = r _i * (R _{ref, max -R ref, min} ) so that now also the measured values between different patterns are comparable. At the same time, the reference values serve to estimate initial exposure parameters for the measurement of all other patterns and thus speed up the measurement.

The final measured values are converted from the camera color space into the XYZ color space. This is done with a camera-specific 3 × 3 color matrix. Such a conversion also takes place as a partial step in the digital development of raw photo data. The determination of this matrix is not part of this invention, only one such matrix is assumed to be known.

However, the measurements thus provided by the camera can only be considered as an approximation of the XYZ values. Due to imperfections of the color filters used, local distortions to the actual XYZ values result. In order to improve the quality, therefore, a hybrid method with a spectrophotometer or colorimeter can optionally be carried out, wherein a subset S 'of S is additionally carried out by means of ordinary spot measurement. The samples in S 'should again be distributed as evenly as possible over the entire RGB space and allow the space of the measurement results of the camera equalize actual measurement results. The number of samples in S 'again allows a tradeoff between speed and quality. It makes sense to use a few hundred of such reference samples.

embodiment

As an example, a DLP projector that receives RGB image data with 8-bit color depth and has a native resolution of 1280 × 800 pixels. Theoretically, up to 2563 (ca 16.8 million) colors can be displayed. To capture a digital camera with a resolution of 4288 × 2844 pixels (about 12 megapixels) is available, which works with three color filters for red, green and blue in Bayer pattern arrangement. The camera is placed so that about 9 megapixels account for the image area of the display.

Example 1: Measurement of 64 ³ Samples

For the dense measurement, 64 ³ color samples are to be recorded, per channel the intensities are to be recorded starting from 0 with a step size of 4, the maximum intensity is then 252 (the remaining measured data can be extrapolated later). Choosing s = 4 will split the RGB cube into 64 sub-cubes, with each pattern containing 4096 samples. For p and q the result is 65. Due to the resolution, an average of 19 × 11 projector pixels are available per color sample. Taking into account the arrangement of the color filters of the camera, this results in an average of approx. 900 sensor elements for green and approx. 450 sensor elements for red and blue per color sample.

For the reference measurement 128 color samples are created, so that each sample can be repeated 32 times in the pattern. The procedure begins with the acquisition of the first subpattern of the reference pattern, which shows only black. The second subpattern shows a very light gray (252, 252, 252). For the reference pattern, 8 permutations are used to capture 8 color subpatterns. In total, 256 repetitions are available per reference sample, so that the reference samples can be acquired with comparatively high quality.

The measurement of each Subpatterns initially by means of an initial exposure and tries to auszubelichten the red channel. After the exposure time and / or ISO value have been further iteratively reduced until no overexposure has occurred in the red channel, starting from the initial exposure, the ISO value and / or exposure time is increased iteratively until there is no longer any underexposed data in the red channel , Whenever correctly exposed data for any color channel and color sample occurs, an additional repeat photograph is taken to increase the quality of the measurement data. The procedure is repeated in green and blue for the remaining channels.

During processing, the sensor data per sample and per color channel are summarized as a weighted average, so that only 3pq values have to be taken into account and collected per photo. After the exposure of all channels, the combination of all photos takes place. First of all, all measured values of photos with the same exposure time are averaged. Then, the exposure that contains the most correctly exposed samples is searched for and used as the base level. Relative to this, all further exposure steps are added iteratively until all exposure steps have been processed, so that p · q measured values are still available for the entire subpattern per camera color channel.

For each color subpattern, all minimum and maximum reference samples are now determined and compared to the respective measurement data from the reference subpatterns. The minimum reference subpattern is black in this case and hardly produces a dynamic background light, while the maximum is almost white and produces a much higher residual light. The color pattern itself contains color samples that lie in between.

In the example, minimizing with the least squares method for the red channel and the minimum would give the equation f (x) = 1.02 x + 0.04 to add the values t from the reference subpattern to the color subpattern analogously for the maximum f (x) = 0.96 x-0.02, whereby the reference values R _{i, min} and R _{i, max} can be determined for each sample in the pattern and finally the values x ' _i can be determined, which correspond to the measured values correspond to the standardized measuring position in the center of the display. This procedure is repeated for all 8 color subpatterns, so that finally for every reference sample 256 measured values are available via which the arithmetic mean is formed. This finally forms the final reference values R _{ref, min} and R _{ref, max} , with which x '' _i can be calculated for the remaining patterns. The procedure is analogous to the method just described for the reference pattern. In the example, however, a smaller number of permutations are chosen for the remaining patterns, so that only 4 color subpatterns and a total of 6 subpatterns per pattern need to be recorded.

In total, there are 384 subpatterns to be detected + 10 subpatterns for reference. On average, 2 exposure levels per pattern are required, with the repetition of the photos, there are about 1800 photos, which can be made in three to five hours depending on the choice of exposure time.

Finally, to obtain the XYZ values, a transformation of the values from the native color space of the camera is required. An affine transformation with a 3 × 3 matrix is used for this purpose.

Example 2: Quick measurement of 17 ³ samples

Analogously to Example 1 should now be made only a quick measurement with significantly reduced sample density, the intensities are to be detected in steps of 16, for the last stage results instead of 256 (which lies outside the capabilities of the display) 255, so that the full color space with 17 ³ = 4913 samples to be detected. It should be noted that the number of samples referred to herein as "reduced" is above what can be measured by conventional methods in a reasonable amount of time.

For the measurement, s = 2 is selected so that only 8 patterns are created, containing a maximum of 729 samples. For p and q, 65 is still selected so that each sample can be repeated at least 5 times in a subpattern. The reference pattern requires only 16 different samples, allowing a 256-fold repetition in a single subpattern. It is therefore sufficient to reduce the number of permutations for the reference pattern to 4, and for the remaining patterns to 2, in order nevertheless to obtain a high number of measurement data at different positions per sample to be measured.

The measuring procedure itself runs as in Example 1. In this case, there are only 6 + 32 subpatterns to be detected. Due to the slightly higher dynamic range of the subpattern arise on average z. For example, you now have 2.5 exposure levels per subpattern, which requires approximately 190 photos that can be captured in 20 to 35 minutes.

bibliography

• [1] International Color Consortium, Specification ICC.1: 2010 (Profile version 4.3.0.0) Image technology color management - Architecture, Profile format, and data structure, 2010 ,
• [2] T. Smith and J. Guild, "The CIE colorimetric standards and their use," Transactions of the Optical Society, Vol. 33, No. 3, pp. 73-134, 1931 ,
• [3] RS Berns, "Methods for characterizing CRT displays," Displays, No. 4, pp. 173-182, May 1996 ,
• [4] T. Siu-Kei, "Color characterization of projectors". USA patent US2006071937 , 2006.
• [5] B. Sajadi, M. Lazarov and A. Majumder, "ADICT: accurate direct and inverse color transformation," in s Proceedings of the 11th European conference on Computer vision: Part IV, Berlin, Springer-Verlag, 2010, pp. 72-86 ,
• [6] PE Debevec and J. Malik, "Recovering High Dynamic Range Radiance Maps from Photographs," Proceedings of the 24th Annual Conference on Computer Graphics and Interactive Techniques SIGGRAPH '97, p. 369-378, 1997 ,

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Sajadi et al. [0012]

Claims (6)

A method for measuring the color outputs of displays with respect to a device-independent color space, depending on the display supplied input values (color samples) by means of a digital camera, characterized in that the amount of color samples to be measured is divided into several patterns and per pattern, two-dimensional reference Subpatterns and several two-dimensional color subpatterns are created so that each reference subpattern consists only of repetitions of a reference color sample derived from the color samples of the pattern, and each color subpattern is a permutation of all the color samples of the pattern as well as repeated repetition of all reference color samples of the pattern Contains patterns so that each subpattern is simultaneously displayed on the display and captured by a set of photos with varying exposure parameters, and the position dependencies of the measurements are determined by relating the measurements of the reference color samples in within the color subpatterns are compensated with the associated measured values of the respective reference subpattern. A method according to claim 1, characterized in that the measured values determined by means of the camera are corrected by a series of comparison measurements with a colorimeter or spectrophotometer. A method according to any one of claims 1 to 2 for the measurement of displays with three-dimensional input color space, characterized in that the division of the color samples to Farbpattern by decomposition of the three-dimensional input color space is carried out in approximately equal partial cubes. A device for measuring the color outputs of displays with respect to a device-independent color space, depending on the display supplied input values (color samples) consisting of a digital camera and one or more computer systems, characterized in that the amount of color samples to be measured is divided into several patterns and creating two two-dimensional reference subpatterns per pattern and a plurality of two-dimensional color subpatterns such that each reference subpattern consists only of repetitions of a reference color sample derived from the color samples of the pattern, and each color subpattern is a permutation of all the color samples of the pattern and the multiple subpatterns Repeat all reference color samples of the pattern so that each subpattern is simultaneously displayed on the display and captured by a set of photos with varying exposure parameters, and the positional dependencies of the measurements through the in-relationship -Setting the measured values of the reference color samples within the color Subpatterns with the associated measured values of the respective reference subpatterns is compensated. A device according to claim 4, supplemented by a colorimeter or spectrophotometer, characterized in that the measured values determined by means of the camera are corrected by a series of comparison measurements with a colorimeter or spectrophotometer. A device according to any one of claims 4 to 6 for the measurement of displays with a three-dimensional input color space, characterized in that the division of the color samples to Farbpattern by decomposition of the three-dimensional input color space is carried out in approximately equal partial cubes.