Scope and Field of Application
PS 3.14 outlines a standardized Display Function for grayscale image representation, offering methods to measure the Characteristic Curve of specific Display Systems This measurement is essential for either adjusting the Display System to align with the Grayscale Standard Display Function or assessing its compliance with this standard Display Systems encompass devices such as monitors with their driving electronics and printers that generate films for light-boxes or alternators.
PS 3.14 is neither a performance nor an image display standard PS 3.14 does not define which
Luminance and luminance range, or optical density range, are essential specifications for image presentation devices However, PS 3.14 does not specify the method for displaying individual picture element values within a particular imaging modality.
PS 3.14 is limited to the display of grayscale images and does not define functions for color image display While Color Display Systems can be calibrated to the Grayscale Standard Display Function for grayscale images, there are no specific requirements for displaying luminance information in color images, regardless of ICC Profile association, on standardized grayscale displays.
Normative References
This Standard includes provisions referenced within the text, which are essential for its application At the time of publication, the listed editions were current; however, all standards are subject to revision Therefore, parties involved in agreements based on this Standard are encouraged to explore the latest editions of the referenced standards for the most up-to-date information.
ISO/IEC Directives, 1989 Part 3 - Drafting and presentation of International Standards.
Definitions
For the purposes of PS 3.14 the following definitions apply.
The Characteristic Curve represents the intrinsic display function of a display system, factoring in ambient light effects It illustrates the relationship between luminance and display drive level (DDL) for emissive devices like CRTs or display controllers, as well as the luminance of reflected light from print media Additionally, it encompasses the luminance derived from optical density measurements of hard-copy media compared to the luminance from a light box This curve is influenced by the operating parameters of the display system.
Note: The Luminance generated by an emissive display may be measured with a photometer Diffuse optical density of a hard-copy may be measured with a densitometer.
Contrast Sensitivity characterizes the sensitivity of the average human observer to Luminance changes of the Standard Target Contrast Sensitivity is inversely proportional to Threshold Modulation.
Contrast Threshold: A function that plots the Just-Noticeable Difference divided by the Luminance across
Digital Driving Level (DDL): A digital value which given as input to a Display System produces a
Luminance refers to the range of discrete values, known as Display Device Levels (DDL), that a Display System can produce The relationship between these DDLs and the resulting Luminance values creates the Characteristic Curve unique to each Display System It's important to note that the output for any specific DDL is determined by the Display System itself and does not account for the Grayscale Standard Display Function.
The Display Function defines how a Display System represents grayscale images by mapping Digital Display Levels (DDLs) to luminance, taking into account the influence of ambient light and the system's adjustment settings This concept is distinct from the Characteristic Curve, which represents the inherent Display Function of the system itself.
Display System: A device or devices that accept DDLs to produce corresponding Luminance values This includes emissive displays, transmissive hardcopy viewed on light boxes, and reflective hardcopy.
Illuminance refers to the ambient light present in the environment around a display system, which plays a crucial role in illuminating the display medium This ambient light affects the luminance perceived by observers, ultimately impacting image quality Additionally, the presence of ambient light can diminish the contrast in displayed images, making it essential to consider lighting conditions for optimal visibility.
Just-Noticeable Difference (JND): The Luminance difference of a given target under given viewing conditions that the average human observer can just perceive.
JND Index: The input value to the Grayscale Standard Display Function, such that one step in JND Index results in a Luminance difference that is a Just-Noticeable Difference.
Luminance is the luminous intensity per unit area projected in a given direction The Système
Internationale unit (used in PS 3.14) is candela per square meter (cd/m2), which is sometimes called nit Another unit often used is footlambert (fL) 1 fL = 3.426 cd/m2
Luminance Range: The span of Luminance values of a Display System from a minimum Luminance to a maximum Luminance.
P-Values are device-independent values defined within a perceptually linear grayscale space, representing the pixel values after all DICOM-defined grayscale transformations have been applied through the DICOM Presentation LUT These P-Values serve as the input for a Standardized Display System.
Grayscale Standard Display Function: The mathematically defined mapping of an input JND index to Luminance values defined in PS 3.14
A Standardized Display System is a device that generates luminance values corresponding to input P-Values through the Grayscale Standard Display Function While the specific method of achieving this is not defined, it can involve transforming P-Values into Display Device Levels (DDLs) that are compatible with the display system.
Standard Luminance Level: Any one of the Standard Luminance levels in Table B-1.
The Standard Target consists of a 2-degree by 2-degree square featuring a horizontal or vertical grating, characterized by a sinusoidal modulation of 4 cycles per degree This square is set against a uniform background that matches the mean luminance of the target.
Note: The Standard Target is defined in terms of the subtended viewing angle, not in terms of the distance from the viewer to the target
Threshold Modulation refers to the minimum level of luminance modulation necessary for the average human observer to detect a Standard Target at a specific mean luminance level This modulation aligns with the Just-Noticeable Difference in luminance of the Standard Target.
Symbols and Abbreviations
The following symbols and abbreviations are used in PS 3.14.
ACR American College of Radiology
ANSI American National Standards Institute
CEN TC251 Comite' Europeen de Normalisation - Technical Committee 251 - Medical Informatics DICOM Digital Imaging and Communications in Medicine
IEEE Institute of Electrical and Electronics Engineers
JIRA Japan Industries Association of Radiological Systems
NEMA National Electrical Manufacturers Association
Conventions
The following conventions are used in PS 3.14:
The terminology defined in Section 3 above is capitalized throughout PS 3.14.
Overview
GENERAL FORMULAS
The Grayscale Standard Display Function is defined by a mathematical interpolation of the 1023
Barten’s model provides a method for calculating luminance levels using the Grayscale Standard Display Function This function enables the determination of luminance, L, in candelas per square meter, based on the Just-Noticeable Difference (JND) Index, j.
L j a c Ln j e Ln j g Ln j m Ln j b Ln j d Ln j f Ln j h Ln j k Ln j
+ ⋅ + ⋅ + ⋅ + ⋅ + ⋅ with Ln referring to the natural logarithm, j the index (1 to 1023) of the Luminance levels Lj of the JNDs, and a = -1.3011877, b = -2.5840191E-2, c = 8.0242636E-2, d = -1.0320229E-1, e = 1.3646699E-1, f 2.8745620E-2, g = -2.5468404E-2, h = -3.1978977E-3, k = 1.2992634E-4, m = 1.3635334E-3.
The logarithms to the base 10 of Luminance Lj are effectively interpolated across the entire Luminance Range, with a maximum relative deviation of 0.3% from the function and a root-mean-square error of just 0.0003 This continuous representation of the Grayscale Standard Display Function enables users to calculate discrete Just Noticeable Differences (JNDs) for any starting levels within the specified Luminance Range.
To effectively apply the formula to a device with a specific range of L values, it is essential to have the inverse relationship, represented as: j L A B Log L C Log L D Log L E Log L.
8 where Log 10 represents logarithm to the base 10, and A = 71.498068, B = 94.593053, C = 41.912053,
2 When incorporating the formulas for L(j) and j(L) into a computer program, the use of double precision is recommended.
3 Alternative methods may be used to calculate the JND Index values One method is use a numerical algorithm such as the Van Vijngaarden-Dekker-Brent method described in Numerical Recipes in C (Cambridge University press, 1991) The value j may be calculated from L iteratively given the Grayscale Standard Display Function’s formula for L(j) Another method would be to use the Grayscale Standard Display Function’s tabulated values of j and L to calculate the j corresponding to an arbitrary L by linearly interpolating between the two nearest tabulated L,j pairs.
4 No specification is intended as to how these formulas are implemented These could be implemented dynamically, by executing the equation directly, or through discrete values, such as a LUT, etc.
Annex B presents the calculated Luminance levels for the 1023 integer JND Indices, while Fig 7-1 illustrates the Grayscale Standard Display Function The specific Luminance levels are influenced by the initial value of 0.05 cd/m².
The Characteristic Curve of a Display System illustrates the relationship between Luminance and Display Drive Level (DDL), factoring in the influence of ambient Illuminance This curve is derived from measurements taken using Standard Test Patterns Specifically, it encompasses the Luminance for emissive displays, such as CRT monitors, in relation to DDL; the Luminance of a transmissive medium positioned before a light-box, influenced by optical density; and the Luminance of a diffusely reflective medium under office lighting, shaped by reflective density, all as functions of DDL.
By internal or external means, the system may have been configured (or calibrated) such that the Characteristic Curve is consistent with the Grayscale Standard Display Function
Certain display systems can adjust to ambient light conditions, typically conforming to the Grayscale Standard Display Function for a specific level of ambient illuminance However, to maintain compliance with this standard, these systems must possess the ability to automatically adjust their display function without requiring user intervention.
TRANSMISSIVE HARDCOPY PRINTERS
For transmissive hardcopy printing, the relationship between luminance, L, and the printed optical density, D, is:
L 0 is the luminance of the light box with no film present, and
L a is the luminance contribution due to ambient illuminance reflected off the film.
If film is to be printed with a density ranging from D min to D max , the final luminance will range between:
L min = L a + L 0 ⋅ 10 − D max , L max = L a + L 0 ⋅ 10 − D min and the j values will correspondingly range from j min = j(L min ) to j max = j(L max )
If this span of j values is represented by an N-bit P-Value, ranging from 0 for j min to 2 N -1 for j max , the j values will correspond to P-Values as follows: j p j p j j
2 1 and the corresponding L values will be L(j(p)).
Finally, converting the L(j(p)) values to densities results in:
Note: Typical values for the parameters used in transmissive hardcopy printing are
REFLECTIVE HARDCOPY PRINTERS
For reflective hardcopy printing, the relationship between luminance, L, and the printed optical density,
L0 is the maximum luminance obtainable from diffuse reflection of the illumination that is present.
If film is to be printed with a density ranging from D min to D max , the final luminance will range between:
L min = L 0 ⋅ 10 − D max , L max = L 0 ⋅ 10 − D min and the j values will correspondingly range from j min = j(L min ) to j max = j(L max )
If this span of j values is represented by an N-bit P-Value, ranging from 0 for j min to 2 N -1 for j max , the j values will correspond to P-Values as follows: j p j p j j
2 1 and the corresponding L values will be L(j(p)).
Finally, converting the L(j(p)) values to densities results in
Note: Typical values for the parameters used in reflective hardcopy printing are
1) Barten, P.G.J., Physical model for the Contrast Sensitivity of the human eye Proc SPIE 1666, 57-
2) Barten, P.G.J., Spatio-temporal model for the Contrast Sensitivity of the human eye and its temporal aspects Proc SPIE 1913-01 (1993)
Fig 7-1 The Grayscale Standard Display Function presented as logarithm-of-Luminance versus
Annex A (INFORMATIVE) A DERIVATION OF THE GRAYSCALE STANDARD
A.1 RATIONALE FOR SELECTING THE GRAYSCALE STANDARD DISPLAY FUNCTION
The Grayscale Standard Display Function was designed to utilize a single continuous and monotonically behaving mathematical function across the entire Luminance Range of interest To simplify its implementation, it was defined using a single data table of pairs Additionally, the function aims to ensure consistent grayscale rendition across displays.
Systems of different Luminance Range and that good use of the available DDLs of a Display System was facilitated.
Perceptual linearization is seen as a valuable concept for developing a Grayscale Standard Display Function to achieve secondary objectives in medical imaging However, it is not an objective in its own right The challenge lies in the difficulty of creating a single mathematical function that can perceptually linearize all types of medical images across diverse viewing conditions Furthermore, medical images are typically displayed using application-specific Display Functions that adjust contrast in a non-uniform manner to meet specific clinical requirements.
It is commonly believed that images that have been perceptually linearized across various display systems will appear similar to viewers To accomplish this perceptual linearization, it was necessary to utilize a model that reflects human visual system responses, for which the Barten model was selected.
Early experiments indicated that a Display Function based on Barten's model of human visual system response could achieve a desirable level of contrast equalization and similarity The images used in these experiments included square patterns, the SMPTE pattern, and the Briggs pattern.
The aim was to connect Digital Display Levels (DDLs) of a Display System to a perceptually linear scale to maximize the effective use of available input levels When digitization levels result in luminance or optical density that are perceptually indistinct, they become redundant Conversely, if the levels are excessively spaced, it may lead to visible contours for the observer Therefore, the idea of perceptual linearization was maintained, although it was not considered an ultimate objective.
Grayscale Standard Display Function, but to obtain a concept for a measure of how well these objectives have been met.
Perceptual linearization is primarily effective for simple images, such as square patterns or gratings in uniform surroundings However, the concept of a perceptually linearized Display Function, developed from experiments with these basic patterns, has been successfully applied to more complex images Although it is acknowledged that perceptual linearization cannot account for all details, spatial frequencies, or object sizes simultaneously, it performs reasonably well for frequencies and object sizes close to the peak of human Contrast Sensitivity, even in intricate images.
Limited unpublished experiments suggest that achieving perceptual linearization for specific details in complex images with a broad luminance range and varied surroundings necessitates display functions that are significantly curved in the darker areas Consequently, these display functions for both low-luminance and high-luminance systems may not form a continuous, monotonic relationship This observation may inform the development of the CIELab curve proposed by various standardization groups.
Recent experiments with computed radiographs indicate that a consistent grayscale representation can be achieved across display systems with varying luminance This is possible when applying a specific function alongside log-linear characteristic curves of the displays Consequently, similarity in image presentation can be attained through a straightforward, luminance-independent display function, rather than relying solely on contrast equalization.
While it might have been equally sensible to choose the rather simple log-linear Display Function as a standard, this was not done for the following reason, among others.
High-resolution display systems with substantial video bandwidth are often limited to digitization resolutions of 8 or 10 bits due to technological constraints A significant deviation of the Grayscale Standard Display Function from the Display System's Characteristic Curve leads to reduced utilization of Display Device Levels (DDL) from a perceptual standpoint The Characteristic Curve of CRT display systems exhibits a convex curvature in relation to a log-linear line, aligning more closely with display functions based on human vision models and perceptual linearization than with a traditional log-linear display function.
When application-specific display processes significantly diverge from the Grayscale Standard Display Function, the resulting Display Function may lack adequate similarity In such instances, alternative functions could offer improved similarity results.
A Display Function was developed from Barten's model of the human visual system to create a continuous mathematical function that balances between log-linear response and perceptual linearization in complex images with a wide luminance range While other models of human contrast sensitivity could potentially offer superior functions, they were not assessed in this context The concept of perceptual linearization was incorporated to address the secondary objectives of the Grayscale Standard Display Function, although it was not the primary aim It is acknowledged that more effective functions may exist to fulfill these objectives, and it is believed that any mathematically defined Standard Function can significantly enhance image presentations on Display Systems within communication networks.
A.2 DETAILS OF THE BARTEN MODEL
Barten's model incorporates various factors affecting visual perception, including neural noise, lateral inhibition, photon noise, external noise, and limited integration capabilities Neural noise sets the upper limit for Contrast Sensitivity at high spatial frequencies, while lateral inhibition in ganglion cells reduces sensitivity to low spatial frequencies by subtracting a low-pass filtered signal Photon noise arises from fluctuations in photon flux, pupil diameter, and the eye's quantum detection efficiency According to the de Vries-Rose law, Contrast Sensitivity at low light levels is proportional to the square root of luminance The model assumes a temporal integration time constant of T = 0.1 seconds, though it does not account for temporal filtering effects Additionally, the eye has constraints on spatial integration, defined by a maximum angular size and a maximum number of cycles over which it can process information amidst various noise sources.
The spatial frequency (u) in cycles per degree (c/deg) is influenced by a Gaussian point-spread function that accounts for the optical properties of the eye lens, stray light from optical media, retinal diffusion, and the discrete nature of receptor elements Additionally, spherical aberration (Csph), which is primarily dependent on pupil diameter, plays a significant role The baseline value of σ (σ0) is observed at smaller pupil sizes External noise can arise from display system noise and image noise, while contrast sensitivity exhibits a sinusoidal variation based on the orientation of the test pattern, showing maximum sensitivity at 0 and 90 degrees and minimal sensitivity at other angles.
45 deg The difference in Contrast Sensitivity is only present at high spatial frequencies The effect is modeled by a variation in integration capability.
The combination of these effects yields the equation for contrast as a function of spatial frequency:
The impact of noise is represented in the initial parenthesis of the square root, highlighting a noise contrast associated with the variances of photons, filtered neurons, and external noise Illuminance plays a crucial role in this context.
The illuminance (IL) of the eye is calculated using the formula IL = π/4 d²L, where d represents the pupil diameter in millimeters and L denotes the luminance of the target in candela per square meter (cd/m²) The pupil diameter can be determined through the equation proposed by de Groot and Gebhard: d = 4.6 - 2.8 tanh(0.4 Log10(0.625 L)).
GENERAL CONSIDERATIONS REGARDING CONFORMANCE AND METRICS
To demonstrate conformance with the Grayscale Standard Display Function is a much more complex task than, for example, validating the responses of a totally digital system to DICOM messages.
Display systems generate analog output, either as luminance or optical density This output can be influenced by various imperfections beyond the inherent flaws in the system's display function that needs validation For instance, spatial non-uniformities in the final image—such as those caused by film, printing, or processing inconsistencies in hardcopy printers—can be measured However, these imperfections typically occur at low spatial frequencies and do not usually impact image quality in diagnostic radiology.
CRTs and light-boxes exhibit inherent spatial non-uniformities that are not addressed by the Grayscale Standard Display Function or the measurement procedures outlined Consequently, even if a test image adheres perfectly to the Grayscale Standard Display Function, its perception on actual CRTs or light-boxes may still be compromised.
The question of how close a display function needs to be to the Grayscale Standard Display Function remains unresolved This uncertainty arises from the lack of psychophysical studies that would identify the minimal difference in display function that is perceptible to an observer when viewing two nearly identical images, such as films shown side-by-side on equivalent light boxes.
The assessment of a Display System can be conducted through visual tests, such as evaluating the perceived contrast of various low-contrast targets within test images, or through quantitative analysis using measured data from instruments like photometers and densitometers.
The quantitative approach can be explored in various ways, such as overlaying plots of measured and theoretical analog outputs, like luminance or optical density, against P-Value, while including error bars to represent expected uncertainties in the measurements A more sophisticated method involves using all measured data points for a statistical analysis to ascertain the underlying Display Function of the Display System, ultimately producing quantitative metrics that evaluate the system's adherence to the Grayscale Standard Display Function.
This article discusses a metric analysis utilizing the "FIT" test to validate the shape of the Characteristic Curve and the "LUM" test to assess the deviation from the ideal Grayscale Standard Display Function This method has been effectively applied to quantitatively demonstrate improvements in the Display Function of specific Display Systems.
This metric approach is presented as one of several options for evaluating Display Systems, rather than the definitive method It is important to take into account specific considerations before choosing or interpreting results from any metric approach.
Practical limitations can restrict the meaningful application of P-Values in analysis, such as when measuring all 256 possible Luminances from a fixed source.
A 12-bit film printer theoretically supports 4096 densities, but practical limitations arise from the accuracy constraints of densitometers and film digitizers Unlike CRT photometers, film density measurements are taken from various locations on the display, leading to spatial non-uniformities that impact hardcopy results Both current hardcopy printers and densitometers exhibit optical density accuracy limitations that are significantly greater than the changes caused by the least significant bit in a 12-bit P-Value While increasing the number of P-Values can potentially capture more localized aberrations from the Grayscale Standard Display Function, it simultaneously reduces the signal-to-noise ratio, diminishing the significance of these measurements.
When dealing with a Display System that exhibits significant measurement noise, characterized by limited repeatability across multiple data sets, it is crucial to apply advanced statistical analysis techniques that incorporate known standard deviations alongside the data This approach helps mitigate the impact of noise on the fitting process, preventing it from overreacting to inaccuracies For further insights, refer to the "General Linear Least Squares" section in Reference C1 and the "Least-Squares Fit to a Polynomial" chapter in Reference C2 Failing to account for measurement noise can result in a misleadingly high root-mean-square error, as it combines errors from both the Display Function's inaccuracies and the measurement noise itself.
To ensure accuracy, it is essential to evaluate the sensitivity and specificity of the metric in question by comparing it with visual tests For instance, utilizing a digital test pattern featuring numerous low-contrast steps under various ambient conditions can provide valuable insights.
Luminances can be printed on a "laboratory standard" Grayscale Standard Display Function printer and compared with prints from an evaluated printer using light-boxes An effective metric technique should sensitively and consistently identify any deviations from the Grayscale Standard Display Function, mirroring human observation For instance, if a Display System's Characteristic Curve exhibits excessive contrast, flatness, or non-monotonic behavior over even a brief range of DDL values, the metric must detect and react to these anomalies with the same sensitivity as a human observer.
In addition to the non-repeatability issues observed in Display System data, it is important to explore other potential sources of variation For instance, altering the order of P-Values in test patterns—whether temporally for CRTs or spatially for printers—could influence outcomes Additionally, using different media with printers may also impact results Therefore, greater confidence can be attributed to metrics that demonstrate stability despite these variations.
METHODOLOGY
To accurately determine the Characteristic Curve of the test Display System, it is essential to conduct numerous measurements as outlined in Sections D.1, D.2, and D.3 The Grayscale Standard Display Function is utilized to calculate the fractional number of Just Noticeable Differences (JNDs) for each luminance interval across evenly spaced P-Value steps JNDs per luminance interval can be derived either directly or through iterative methods, such as linear interpolation when only a few JNDs are associated with each interval Following the transformation of the display system's grayscale response, the luminance levels for each P-Value are denoted as Li, with the corresponding standard luminance levels represented as Lj The variable dj indicates the JNDs per luminance interval on the Grayscale Standard Display Function for the specified number of P-Values Consequently, the JNDs per luminance interval for the transformed display function can be calculated using the formula r = dj (Li+1 - Li)(Lj+1 + Lj) / ((Li+1 + Li)(Lj+1 - Lj)).
An iterative method can effectively calculate the number of Just Noticeable Differences (JNDs) within a specific Luminance interval by utilizing the Grayscale Standard Display Function This approach involves counting the complete JND steps within the interval and assessing any remaining fractional step To begin, start at the lower end of the Luminance interval and determine the Luminance step corresponding to one JND Continue this process until reaching the upper end of the Luminance Range, then calculate the fractional portion represented by the last step The final result combines the total number of completed integer JND steps with the fractional portion of the last step, yielding the total number of JNDs in the specified Luminance interval.
The Luminance intervals vs JNDs curve illustrates the relationship between the number of Just Noticeable Differences (JNDs) and the index of luminance intervals, with JNDs plotted on the vertical axis and luminance interval indices on the horizontal axis As demonstrated in figure C-1, a linear regression analysis reveals that the data points align closely with a horizontal line, indicating a strong correlation between luminance intervals and JNDs.
Figure C-1 Illustration for the LUM and FIT conformance measures
The evaluation of JNDs/Luminance interval data is conducted using two statistical measures The first measure assesses the overall compatibility of the test Display Function with the Grayscale Standard Display Function, while the second measure focuses on the local approximation of the Grayscale Standard Display Function to the test Display Function.
After verifying the assumptions of normal multiple linear regression, two key measures are utilized in the analysis The first, known as the FIT test, compares the Luminance-Intervals-vs-JNDs curve of the test Luminance distribution with various polynomial fits The ideal Grayscale Standard Display Function exhibits a constant number of JNDs per Luminance interval across the entire range, represented by a horizontal line This indicates that both local and global means of JNDs per Luminance interval remain constant If the data aligns more closely with a higher-order polynomial, it suggests that the distribution does not accurately reflect the Grayscale Standard Display Function, necessitating further regression analysis using third-order curves for comparison.
The Luminance Uniformity Metric (LUM) evaluates the uniformity of luminance steps in perceptual size (JNDs) across the luminance range This assessment is quantified using the Root Mean Square Error (RMSE) of the curve fitted by a horizontal line representing the JNDs/luminance interval A lower RMSE indicates that the test display function closely matches the Grayscale Standard Display Function on a microscopic level.
Both the FIT and LUM measures can be conveniently calculated on standard statistical packages
If the luminance distribution successfully passes the FIT test, its quality is assessed using a single quantitative measurement known as LUM, which represents the standard deviation of the just noticeable differences (JNDs) within the luminance interval from their mean It is anticipated that clinical practice will establish the acceptable tolerances for both FIT and LUM values.
To achieve a close approximation of a test Display Function to the Grayscale Standard Display Function, the number of discrete output levels in the Display System is crucial For example, enhancing the LUM measurement can be accomplished by utilizing a subset of available DDLs, which preserves the full output digitization resolution but may reduce contrast resolution.
The LUM is affected by the selection of discrete output gray levels in the Grayscale Standard Display Function The optimal number of output levels is dictated by the specific clinical application, which may involve independent grayscale image processing.
The Grayscale Standard Display Function (GSDF) standardization, outlined in PS 3.14, does not specify a fixed number of gray levels for output However, a greater number of distinguishable gray levels typically enhances image quality by improving contrast resolution It is advisable to determine the necessary output driving levels for the transformed Display Function before standardizing the Display System, based on its clinical applications This approach ensures that the transformation calculations utilize an appropriate gray scale distribution, avoiding insufficient output levels.
EMISSIVE DISPLAY SYSTEMS
D.1.1 Measuring the system Characteristic Curve
Before measuring the Luminance response of an emissive Display System, it is essential to allow the device to warm up according to the manufacturer's recommendations and to adjust it in line with their performance specifications Specifically, adjustment procedures for setting the black and white levels should be sourced from the manufacturer The objective is to optimize the dynamic Luminance Range while avoiding artifacts, thereby achieving the maximum number of Just-Noticeable Differences (JNDs).
To ensure the system is correctly configured, a straightforward test can be conducted by examining the SMPTE pattern The contrast perceived between the 5% square and its surrounding 0% area should match the contrast perceived between the 95% square and a white square.
To measure the Characteristic Curve of a Display System, utilize a test pattern that features a central square measurement field, which constitutes 10% of the total pixel count, surrounded by a full-screen uniform background set at 20% of the maximum luminance.
The measurement field of a CRT monitor is set at 10% of the total displayed pixels, with a surrounding area configured to 20% of the maximum luminance This setup leads to internal light scatter, resulting in a luminance range that is typically comparable to that observed in radiographs, such as thorax radiographs.
Figure D.1-1 The test pattern will be a variable intensity square in the center of a low Luminance background area
Notes: 1 For example, on a 5-megapixel Display System with a matrix of 2048 by 2560 pixels, the target would be a square with 724 pixels on each side.
For optimal results, the test pattern should occupy the full screen However, in some windowed operating environments, it can be challenging to remove certain user-interface elements, such as menu bars at the top In these situations, the background should cover as much of the screen as feasible.
The Characteristic Curve of the Display System may be determined by
- turning off all ambient lighting (necessary only when a suction cup photometer is used or when a handheld photometer casts a shadow on the display screen);
- displaying the above test pattern;
- setting the DDL for the measurement field to a sequence of different values, starting with 0 and increasing at each step until the maximum DDL is reached;
- using a photometer to measure and record the Luminance of the measurement field at each command value.
To accurately model the Characteristic Curve of the Display System, it is essential to have an adequate number and strategic distribution of DDLs for measurements, as outlined in Annex C.
Notes: 1 If a handheld photometer is used, it should be placed at a distance from the display screen so that
Luminance is assessed at the center of the measurement area, ensuring that it does not overlap with the surrounding background The distance for this measurement can be determined using the acceptance angle specifications provided by the photometer's manufacturer.
The precise number and arrangement of DDLs should be determined by the Display System's features and the mathematical method employed to interpolate its Characteristic Curve It is advisable to utilize a minimum of 64 distinct command values in this process.
To ensure accurate Luminance measurements, it is crucial to space them appropriately over time, allowing the Display System to reach a steady state This is especially important for the initial measurement at DDL 0, where giving the system time to settle can significantly impact the results.
When configuring a Display System to adhere to the Grayscale Standard Display Function, it is essential to consider the impact of ambient light on the apparent Characteristic Curve.
Using a handheld photometer that eliminates shadowing on the display screen allows for simultaneous measurement of the Characteristic Curve, capturing both the Luminance generated by the display and the influence of ambient light.
To accurately measure the Characteristic Curve using a suction cup or handheld photometer, all ambient lighting must be turned off during the measurement process Ambient light's effect is assessed separately by turning off the Display System and measuring the Luminance from ambient light scattering on the display screen This is done by positioning the photometer at an appropriate distance to capture a significant portion of the screen while avoiding direct light interference The Luminance from ambient light is then combined with the previously measured levels from the Display System to establish the effective Characteristic Curve of the system.
Note: Changes in ambient lighting conditions may require recalibration of the display subsystem in order to maintain conformance to this standard.
This article presents an example of measurements and transformations of a Display Function using a CRT monitor equipped with a display controller The display controller is capable of transforming Display Device Levels (DDL) with 8-bit input precision to 10-bit output precision.
The Luminance is measured with a photometer with a narrow (1o) acceptance angle The ambient light level was adjusted as low as possible No localized highlights were visible.
The maximum Luminance was determined by adjusting the DDL of the measurement field to its peak value, while the DDL of the surrounding area was set to the midpoint of its range Consequently, the Luminance for the surrounding area was calculated as 20% less than the maximum Luminance obtained from the measurement field.
The ambient light was disabled, and the photometer was positioned at the center of the measurement field for the test pattern shown in Fig D.1-1 Luminance measurements were taken while adjusting the input level Dm in increments.
TRANSPARENT HARDCOPY DEVICES
D.2.1 Measuring the system Characteristic Curve
A laser printer serves as a transparent hardcopy device, utilizing a processor to print multiple images onto a sheet of transparent film, usually sized 14” x 17” This film is then positioned over a high luminance light box in a darkened environment for optimal viewing.
The characteristic curve for a transparent hardcopy device is established by printing a test image featuring a series of n bars, each assigned a specific numeric value (DDL) The optical density of each printed bar is measured with a transmission densitometer, providing essential data for analysis.
To accurately define a printer's Characteristic Curve, it is beneficial to maximize the number of data points (n) captured However, limitations in quantitative repeatability due to printer, processor, or media technologies may necessitate a smaller value of n This adjustment helps maintain a stable and meaningful conformance metric, as excessive density differences between adjacent bars can fall within the noise level when too many bars are included.
One example of a test image is a pattern of 32 approximately equal-height bars, spanning the usable printable region of the film, having 32 approximately equi-spaced DDLs as follows:
The article outlines a comprehensive guide on density, structured in a step-by-step format from Density Step 0 to Density Step 31 Each step provides essential insights into understanding density, allowing readers to grasp its fundamental concepts progressively This systematic approach ensures clarity and enhances learning, making it an invaluable resource for anyone seeking to deepen their knowledge of density.
Figure D.2-1 Layout of a Test Pattern for Transparent Hardcopy Media
To define a test pattern with n DDLs for a printer with an N-bit input, the DDL of step # i can be set to
DDLi = (2 N -1) (D.2-1) rounded to the nearest integer.
The tabulated values of DDLi and the corresponding measured optical densities ODi constitute a
Characteristic Curve of the printer.
D.2.2 Application of the Grayscale Standard Display Function
Films produced by transparent hardcopy printers are frequently displayed in various locations, allowing for viewing on different light boxes and under diverse lighting conditions.
The PS 3.14 standard specifies the required densities for hardcopy transparent printers, focusing on producing accurate grayscale images It outlines a method for applying the Grayscale Standard Display Function specifically to transparent hardcopy outputs, utilizing parameters that reflect typical light-box luminance and viewing conditions.
The specific parameters which are used in the following example are as follows:
L0 (Luminance of light-box with no film present): 2000 cd/m2
La (ambient room light reflected by film): 10 cd/m2
Dmin (minimum optical density obtainable on film): 0.20
Dmax (maximum optical density desirable on film): 3.00.
The process of constructing a table of desired OD values from the Grayscale Standard Display Function begins with defining the Luminance Range and the corresponding range of the Just-Noticeable
Difference Index, j The minimum and maximum Luminance values are given respectively by
Next, calculate the corresponding Just-Noticeable Difference Index values, jmin and jmax For the current example, we obtain j min = 233.32 (D.2-4) j max = 848.75 (D.2-5)
The printer must establish a range of j-values to ensure proper functionality, mapping the minimum input (P-Value = 0) to jmin and its corresponding Lmin Additionally, the printer should also map its maximum input (P-Value = 2N) to the appropriate jmax value.
1 where N is the number of input bits) to jmax and the corresponding Lmax At any intermediate input it should map its input proportionately: j (PV) = j min + (j max -j min ) PV
To achieve the desired luminance L(j(P-Value)) as specified by the standard's formula, it is essential to produce an optical density (OD) that corresponds to the P-Value This targeting process ensures that the calculated density meets the conditions established for L0 and La The necessary density can be accurately determined through a specific calculation.
D.2.3 Implementation of the Grayscale Standard Display Function
In the case of a printer utilizing an 8-bit input, a table can be established to define the output densities (ODs) corresponding to each of the 256 possible P-values.
Table D.2-1 Optical Densities for Each P-Value for an 8-Bit Printer
Plotting these values gives the curve of Figure D.2-3.
Figure D.2-3 Plot of OD vs P-Value for an 8-Bit Printer
As an example, a bar pattern with 32 optical densities was printed on transmissive media (film)
The printer was initially configured to operate within a density range of 0.2 (Dmin) to 3.0 (Dmax) and was pre-set by the manufacturer to utilize the Grayscale Standard Display Function This function was subsequently transformed into a table that correlates target density values with P-Values, as previously outlined.
The test pattern which was used for this was an 8-bit image consisting essentially of 32 horizontal bars The 32 P-Values used for the bars were as follows: 0, 8, 16, 25, 33, 41, 49, 58, 66, 74, 82, 90,99, 107,
In this study, the optical densities of 32 bars in a film were measured and converted to luminance using standard light-box parameters and reflected ambient light These luminance values were then transformed into Just-Noticeable Difference (JND) indices by applying a mathematical function that relates luminance to JND For each of the 31 intervals between consecutive measurements, the "JNDs per increment in P-Values" was calculated by dividing the difference in JND indices by the difference in P-Values for that interval Notably, all calculations involving density, luminance, and JND were performed using floating-point variables to avoid truncation errors.
In this analysis, the film's data can be effectively represented by a horizontal straight line, indicating that the calculated "Just Noticeable Differences" (JNDs) per increment in P-Values remains consistently at 2.4 A mathematical fitting of the data supports this conclusion.
The total difference of 0.2 is significantly smaller than the noise observed in the 31 individual values of "Just Noticeable Differences (JNDs) per increment in P-Value," raising doubts about its significance This noise represents the random, non-repeatable variations that occur when comparing new sets of measured data, such as from a second print of the same test pattern, with previous measurements.
No visual tests were done to see if a slope that small could be detected by a human observer in side-by- side film comparisons.
When analyzing the 32 original absolute measured densities, there is a notable alignment between the target and measured optical densities, adhering to the manufacturer's accuracy standards However, when employing metrics based on differential information over small intervals, the results warrant careful consideration, as they can be significantly influenced by various imperfections that do not reflect the device's true or averaged performance characteristics.
REFLECTIVE DISPLAY SYSTEMS
This example demonstrates how adherence to the Grayscale Standard Display Function can be achieved using a thermal-dye-transfer paper printer combined with an office-light system The printer generates black-and-white grayscale images on semi-glossy, heavy-gauge 8-inch x 10-inch paper, which is evenly illuminated by fluorescent lamps, resulting in a minimum reflective density that yields a luminance of 150 cd/m² It is assumed that the transformation operator maintains an equal input and output digitization resolution of 8 bits.
D.3.1 Measuring the system Characteristic Curve
A 64-step grayscale tablet was printed for DDLs ranging from 4 to 255, and the reflection optical densities were measured using a densitometer, yielding values from 0.08 to 2.80 The corresponding luminance levels for these optical densities under various illumination conditions are illustrated in Fig D.3-1.
Figure D.3-1 Measured and interpolated Characteristic Curve and Grayscale Standard Display
Function for a printer producing reflective hard-copies
D.3.2 Application of the Grayscale Standard Display Function
This example demonstrates how to achieve compliance with the Grayscale Standard Display Function using a thermal-dye-transfer printer and an office-light system The printer generates black-and-white grayscale images on semi-glossy, heavy-gauge 8-inch x 10-inch paper, which is evenly illuminated by fluorescent lamps to ensure optimal reflective density.
Luminance of 150 cd/m 2 The hypothetical transformation operator is assumed to have equal input and output digitization resolution of 8 bits.
D.3.3 Implementation of the Grayscale Standard Display Function
The measured Characteristic Curve is interpolated for the available DDLs yielding 256 Luminance levels
L I,m The Grayscale Standard Display Function is also interpolated between JND min and JND max (DJND = [JND max - JND min ]/255) yielding 256 Standard Luminance levels L I,STD
For every L I,STD , the closest L J,m is determined The data pair I,J defines the transformation between
D input and D output (Table D.3-1 and Fig D.3-2) by which the Luminance response of the Display System is made to approximates the Grayscale Standard Display Function.
Figure D.3-2 Transformation for modifying the Characteristic Curve of the printer to a Display
Function that approximates the Grayscale Standard Display Function
Table D.3-1 Look-Up Table for Calibrating Reflection Hardcopy System
P-Value DDL P-Value DDL P-Value DDL P-Value DDL
The FIT and LUM metrics outlined in Annex C are utilized to assess both macroscopic and microscopic approximations of the LJ,m to the LI,STD Figure D.3-3 illustrates the perceptually linearized Display Function overlaid on the Grayscale Standard Display Function, while Figure D.3-4 summarizes the results from the two metrics The FIT metric indicates a strong global fit, presenting a straight line closely aligned with the horizontal axis for the JNDs/Luminance interval Conversely, the LUM metric reveals a relatively high RMSE, signifying notable local deviations from the Grayscale Standard Display Function, particularly evident in the soft-copy Display System discussed in Section D.1 This increased RMSE partly results from the equal input and output digitization resolution during transformation Additionally, the transformation table (Table D.3-1) and Figure D.3-2 demonstrate that multiple P-Values correspond to the same Luminance levels in the transformed Display Function, with only 205 of the 255 Luminance intervals yielding JNDs for the Standard Target.
The transformed Display Function for a reflective hard-copy display system aligns perfectly with the Grayscale Standard Display Function, resulting in two superimposed and indistinguishable curves.
Figure D.3-4 Measures of conformance for a reflective hard-copy Display System with equal input and output digitization resolution of 8 bits
Annex E (INFORMATIVE) REALIZABLE JND RANGE OF A DISPLAY UNDER
Dynamic range is a crucial measure of the information capacity of a display system, yet it is often poorly defined Many definitions, such as Poynton's contrast ratio, focus on ideal conditions, measuring the intensity difference between the brightest white and the darkest black However, real-world factors like veiling glare, noise, spatial frequency, power supply saturation, and ambient lighting significantly affect the actual dynamic range experienced in practical scenarios, particularly in cathode ray tube (CRT) displays Consequently, the concept of dynamic range remains ambiguous and challenging to quantify accurately in display systems.
Note: Veiling Glare is the phenomenon wherein internal light reflections inside the CRT creates a
"background lighting" thus reducing the contrast range of the CRT device.
The techniques employed to assess how closely a Display System's Display Function aligns with the Grayscale Standard Display Function can also establish two key metrics that effectively characterize a Display System's ability to convey information These metrics, known as theoretically achievable Just Noticeable Differences (JNDs) and realized JNDs, are valuable for comparing the performance of different Display Systems.
The theoretically achievable Just Noticeable Differences (JNDs) in a display system are determined by the visual model based on the system's luminance range To find the number of JNDs, refer to Table B-1 in Annex B and count the JNDs that lie between the system's measured minimum and maximum luminance levels.
The achievable number of Just Noticeable Differences (JNDs) in a Display System may be limited by the resolution capabilities of its components, particularly the quantization resolution determined by the finite bits per pixel For instance, a Display System might theoretically deliver 352 JNDs, as indicated in Table B-1 of Annex B However, if the system operates with only 8 bits per pixel, the maximum attainable JNDs would be restricted to 256 due to the quantization effect.
The number of Just Noticeable Differences (JNDs) in a Display System is always less than or equal to the minimum of the theoretically achievable JNDs and the quantization limit This limitation arises because some quantized values sent to the display may not align with the input values necessary to reach the subsequent JND.
The useful number of realized Just Noticeable Differences (JNDs) quantifies the actual JNDs achieved by a Display System, considering factors like contrast resolution and luminance distribution This metric reflects the informational dynamic range that the system can effectively convey to a human observer To calculate this number, one starts at the minimum luminance and increments by one JND in luminance, selecting the smallest change in Display Device Level (DDL) that meets or exceeds this increment By systematically applying this process across all available DDLs, a sequence of steps, each at least one JND apart, is generated, with the total length of this sequence representing the number of realizable JNDs for the Display System.
The PS 3.14 methods may not perfectly replicate every real-world source of degradation in a Display System, yet they provide a consistent approach to assess the achievable number of Just Noticeable Differences (JNDs) This uniformity offers valuable insights into the actual performance of a Display System, reflecting the experience of a human observer in practical scenarios, such as evaluating radiological images in the medical field.