Purpose:
Contrast has long been a key metric when discussing the image quality generated by video displays, but in this era of dynamic irises, fixed irises, native only, high ANSI contrast but average on/off and high on/off but average ANSI contrast a lot of confusion exists over the pros and cons of different display technologies from a contrast perspective. This project was started as a way to provide a consistent and measurable contrast benchmark that can be applied equally to all display technologies to help understand their contrast strengths and weaknesses. To this end, two sets of arbitrarily generated test patterns are provided which can be used as a framework for analyzing the relative differences in intra-image contrast for both native and dynamic iris projectors alike. Two test patterns were used so that the effects on contrast from both static and dynamic mechanisms can be fully captured.
The hope is that AVS members will use these included test patterns and supplied methodology to measure their display devices and add them to the AVS contrast database as an ongoing effort (hence the reason this thread was made "sticky"). The pages below provide a technical discussion of this contrast benchmark and how it can be applied to video projectors. Two off-the-shelf projectors were used for this example discussion, one is a native projector (JVC RS-1) and the other is a projector that utilizes a dynamic iris (Sony VW50 Pearl). At the end of this discussion is a section that contains the actual ongoing AVS Contrast project data and results which will be continually updated as information is provided.
Terminology:
To avoid confusion lets define some key terms which will be encountered in this thread:
Intra-Image Contrast - This is often referred to as simultaneous contrast and it refers to the contrast within regions of one specific image at the same point in time. ANSI contrast is a well known example of intra-image contrast because the black and white measurements are done while both checkerboxes are displayed simultaneously on the screen. As we will see, virtually all forms of contrast measurements in the AVS contrast project are either a direct measurement of intra-image Contrast or extrapolated from intra-image measurements. The term intra-image contrast can apply to any type of an image but in this discussion it is exclusively used to describe the contrast of test patterns and the term intra-scene contrast is used to describe real-world images from video and film sources.
Static Intra-Image Contrast - In this thread we use this term to describe the intra-image contrast created by projectors that don't perform electronic alterations of the image to boost brightness of portions of an image that are not already at full white. Native contrast (i.e. non-dynamic iris) projectors and fixed iris projectors are the most well known examples of projectors of this category. In this discussion however we will also use the term to include the contrast effects from a dynamically changing iris aperture (but not the brightness boost from dynamic gamma). Most (but not all) DI equipped projectors perform some form of gamma boost so the minority that do not can be fully treated using the static intra-image methodology. As we will see, Dynamic iris projectors that utilize gamma boost can and should be measured for static intra-image contrast which is useful for comparison purposes.
Dynamic Intra-Image Contrast - In this thread we use this term to describe the intra-image contrast created by projectors that can selectively boost the brightness of less than 100% full white pixels in some images. This contrast boost improves shadow detail and is performed by boosting the stimulus of pixels (not already at their maximum) which is currently always done (but doesn't have to be) in combination with a reduction in iris aperture. The boosting of the stimulus of pixels that aren't at their maximum is called Dynamic Gamma. Dynamic Iris projectors are the most well known examples of this technology and benefit the most from dynamic intra-image contrast, although as we will see a non-DI equipped projector can also be measured using the same methodology and test patterns as a DI equipped projector which is useful for comparison purposes.
Maximum Dynamic Range - We will use this term to refer to the maximum range of white to black that a projector is capable of achieving but not necessarily at the same time in the same image. This is usually expressed by the sequential on/off contrast of a projector. As we will see, this isn't the same as the maximum intra-image contrast because DI equipped projectors can not achieve full white and dark blacks at the same time in the same image.
Maximum Intra-Image Contrast - This term represents the largest range of black to white contrast that a projector is capable of achieving in an image at one time. As contrast is a ratio of white to black, determining the maximum intra-image contrast entails using a 100% stimulus white reference level to ascertain the maximum intra-image contrast. As we will show the on/off contrast of a native projector provides a surprisingly close approximation to the maximum intra-image contrast. DI projectors on the other hand do not achieve full white and full black simultaneously in the same image so mechanisms other than simple sequential on/off contrast measurements are needed in order to determine both their maximum intra-image contrast as well as their dynamic contrast benefits.
Intra-Scene Contrast - When used in this thread this term refers to intra-image contrast in video or film sources rather than the images used in a specific test pattern to determine intra-image contrast. Intra-Scene contrast is complicated by the luminance distribution of the image, the spatial relationships between light and dark pixels as well as the gamma used by a projector. Because of these complications, it must be stressed that intra-image contrast and intra-scene contrast are not the same. Measuring intra-image contrast is still useful in providing a basis of comparison between projectors and higher contrast results from this intra-image benchmark generally equates to improvements in intra-scene contrast in much the same way that high on/off contrast equates to improvements in low APL images and improvements in ANSI contrast equates to improvements in higher APL images. Despite the complexities of Intra-Scene contrast, it is possible to compare specific regions of well understood scenes (where both the luminance and spatial relationships are known) and determine roughly how various intra-image contrast metrics may affect these scenes (see this cine4home article for example:
http://www.cine4home.de/Specials/ANS...NSIvsONOFF.htm ).
Background:
Traditionally two forms of contrast measurements have been used to measure contrast. Sequential On/Off contrast uses alternating full fields of 100% white and 0% black to determine the maximum brightness and darkness that a display device is capable of achieving. ANSI contrast on the other hand uses a 4x4 checkerboard pattern of full white and full black to determine the contrast achievable when 50% of the image is black and the other white. ANSI contrast is often said to be the only metric of intra-image contrast because its contrast is measured with both the white and dark patterns displayed simultaneously. It's important to realize that this is a myth and that both ANSI and on/off contrast play important roles in intra-image contrast. Dark, low APL images (both test patterns and video/film scenes) are more heavily influenced by the on/off contrast while bright content is more heavily influenced by ANSI contrast. On average, movies are inherently a dark medium so on/off contrast plays a large role although ANSI still plays an important role in bright scenes (and sporting events, etc.).
Dynamic and Static Intra-Image Contrast:
Segregating contrast into dynamic and static contrast categories gives the mistaken impression that there are two unique types of contrast. The reality is that there is only one type contrast and that the terms dynamic and static contrast really refer to the mechansims employed by the display device to render contrast in an image. Display devices with dynamic contrast mechanisms have become common only fairly recently. Determining static contrast parameters and also dynamic contrast parameters with these display devices is required because they use a mix of static and dynamic mechanisms.
In this thread we only look at intra-image contrast and we employ two different techniques and two different sets of test patterns to examine the full benefits of contrast from both static and dynamic mechanisms. Because the same intra-image contrast is being measured in both cases (but with different techniques) there is considerable overlap between both discussions and we will see that both technologies can and should be compared using the same techniques and methodologies.
What causes Intra-Image Contrast?
We all know that light from a lamp or light source creates the whites in an image, but we will also examine the constancy of the brightness of white regions in both native and dynamic iris projectors. The real contributor of display contrast however is the black level in regions of an image. The black level of these regions is the sum of the minimum black level of the device (all pixels off) coupled with variance in dark regions of an image as it is influenced by the luminance in the bright regions. Variations in the amount of luminance (ie more pixels turned on or off) or variations in the intensity (brightness) of the luminance will have profound changes on the black level in regions of an image.
The underlying reason for the variance of intra-image black levels is due to leakage and scatter of light penetrating the dark regions of the image. The factors that affect light leakage and scatter are complex and include factors such as the device/pixel characteristics of the display device, the effectiveness of light traps, polarizors and internal light baffling, the use and degree of internal irises and also optical characteristics such as dispersion and distortion in the lens. These factors all combine to reduce the overall contrast of the system. In this project we address only the system contrast measured at the projector and not contrast effects from individual sources such as the room or screen (although effects from the room and screen may be addressed at a later date).
Intra-Image Contrast as a function of Luminance.
As mentioned the causes of intra-image contrast are complex and caused by many factors. This complexity also affects contrast measurements because both the geometry and overall amount of luminance can play a role. As an example, the washout effect on dark regions from bright regions in an image varies as one moves further away from the bright region. We could for example construct two different test patterns with different geometries but with the same overall luminance and measure different contrast results. If the geometries were kept relatively similar however we can see contrast change as both the overall amount and/or intensity of luminance is changed.
This relationship of contrast vs luminance (both in amount and intensity) is an important concept because it helps us to tie together the intra-image contrast benefits from both on/off contrast and ANSI contrast and it shows that they influence Intra-Image contrast in different regions of the luminance range. By using a fixed set of test patterns (even with an arbitrary geometry) that varies by the amount of luminance, a synthetic benchmark of contrast vs luminance can be obtained. Applying this contrast benchmark to projectors allows us to infer relative performance differences between projectors.
This methodology is similar to measuring and applying traditional ANSI contrast but with the added benefit that the contrast performance in the low luminance range can also be ascertained. As we will see, for DI-equipped projectors, utilizing low luminance contrast test patterns is the only way to gauge intra-image contrast with dark material because the low APL of these patterns allows the dynamic iris to close.
It should be stressed that the shape of the contrast curve is affected by the geometry and the amount and intensity of the luminance of the test patterns used. As we will see, the shape of the contrast curve using the test patterns generated for the AVS contrast project is completely arbitrary but it still follows an inverse power curve relationship between contrast and luminance that was predicted by AVS forum member Erik Garci.
Geometrical differences between each of the test patterns within this suite cause some variation in this relationship but overall the measurements correlate surprisingly well to this relationship. This points out a key point however that the precise shape of the static intra-image contrast curves that will be presented is synthetic and influenced by the arbitrary geometry of the test patterns chosen. This is only one such possible benchmark, but it nonetheless allows us to make some interesting comparisons and to infer relative performance differences versus luminance that we can't do otherwise. It is only one possible benchmark but as far as I know it's the only one of its type in the industry.
As we will see these test patterns and associated methodology also has the surprising benefit of allowing us to determine the maximum intra-image contrast of a display device. The maximum intra-image contrast occurs when the overall amount (area) of 100% full white luminance approaches 0 in an image. Interestingly, we will show that the maximum intra-image contrast is not dependent on the geometry of the test pattern being used but using low APL test patterns is the only reliable way to determine it.
To illustrate what static intra-image contrast vs luminance looks like, the graph below shows a theoretical plot of contrast vs luminance which does not take into account geometric factors and was created using a mathematical model by AVS forum member Erik Garci (An online version of this model can be found here:
http://home1.gte.net/res18h39/contrast.htm ). This model and associated plot shows a classic inverse power function that Erik derived for his model.
As one can see from this graph of theoretical intra-image contrast, traditional ANSI contrast corresponds to the 50% luminance end point and traditional on/off contrast corresponds to the 0% luminance end point.
If we compare this model against the static contrast test pattern suite that we've created we can see a relatively close correlation to this inverse power function relationship even though the test patterns used were constructed independently and without any prior knowledge of the contrast model. The graph below illustrates this comparison. Deviations from this model are likely due to variations in the geometry of each individual test pattern used in the suite and in fact we see this in a slight jaggedness of the curve particularly at the 2% luminance point which also happens to correspond to a test pattern where the geometry happens to place a higher percentage of white content around the dark measurement region.
Rather than plot the x-axis luminance scale linearly as is done above, it's useful to plot the results of each test pattern with an equal spacing so that the results from each test pattern in all luminance ranges can be more easily seen. Plotting the data this way makes it easier to see the full range of intra-image contrast values but it does alter the shape of the curve from the expected inverse power curve relationship. The reader should keep this in mind as nearly all of the graphs and data presented will use this graphing convention even though it alters the shape of the curve.
Test Pattern Suites: Two sets of test patterns have been developed by William Phelps and I for characterizing intra-image contrast. The first set of patterns is designed to measure maximum intra-image contrast and this is accomplished by varying 100% full white luminance by area. Luminance is varied in this way from 50% (modified ANSI) to 0.1% (0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10, 15, 20 and 50%). For each step on the luminance scale, two test patterns are used to determine the white and black levels for that luminance step. Using two test patterns ensures that the probe placement does not change between white and black readings. A separate white reading is also needed at each luminance step because the constancy of the white levels can not be guaranteed (which is particularly true for a dynamic iris).
All of the patterns are designed for center probe placement which does not change which helps to ensure accuracy. The 50% pattern is a modified variation of the ANSI pattern designed for the same center probe position used by the other patterns. Full field white and Full field black patterns are also included so that traditional on/off contrast can be measured at the same time.
One key point about this set of Intra-Image test patterns is that they measure intra-image contrast by utilizing full white which is needed to determine the maximum contrast of a display device. By utilizing 100% full white these results are also not dependent on projector gamma. 100% full white also allows us to determine the maximum intra-image contrast of a display device. Measured in this way we can see the full benefits to absolute intra-image contrast from both dynamic and static display technologies, but we are unable to measure the full benefits of dynamic contrast which occurs with less than 100% stimulus white and which help to improve shadow detail. It was for this reason that the 2nd set of test patterns was developed.
Black Test Pattern for 50% luminance (modified ANSI pattern designed for center probe position)
White Test Pattern for 50% luminance (modified ANSI pattern designed for center probe position
Black and White Test Patterns for 1% luminance range
The second set of test patterns are designed to measure the full benefit of intra-image contrast from display devices that dynamically enhance the contrast (e.g. dynamic contrast). These test patterns achieve this by using the same geometrical pattern but with luminance varied by intensity rather than by area as is done with the static contrast test patterns described above. Using the same white area in each pattern, the intensity of the white regions are varied from 100% down to 1% (1,2, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100).
In this discussion we will use the naming convention of "% pixel stimulus" as a way to describe the way in which discrete changes in pixel brightness are modulated from 0 to 100% rather than the 0 IRE and 100 IRE convention. This was done to avoid confusion between PC/Video levels, 7.5IRE black enhance, etc. It's worth mentioning however that the test patterns were designed with video levels (16-235) in mind and the proper application of video levels (as well as properly set brightness and contrast controls) are critical to ensure accurate results with this set of patterns.
It should also be pointed out that in this discussion we use the term luminance (either relative or absolute) to mean light that that is being generated by the projector in response to the number of pixels of the specified % stimulus.
We should also point out that display gamma plays a role in the brightness of these dynamic contrast test patterns and therefore contrast is also a function of display gamma for these measurements and this will be explored in more detail later.
In order to provide continuity between the static and dynamic suite of test patterns, the same 0.5% (by area) luminance/low APL white pattern from the suite of static test patterns is used for all off the dynamic test patterns. The 0.5% pattern was chosen because it's the largest pattern which reliably fills the probe area while also being small enough to force a dynamic iris projector to use its smallest aperture and presumably the largest gamma boost settings so that the maximum benefit of dynamic contrast can be seen for all luminance ranges from 100% to 1% stimulus.
Both sets of test patterns were designed for ease of use so that any forum member with access to a relatively inexpensive light meter such as an AEMC CA813 will be able to use these patterns to determine the static and dynamic contrast of their projectors and contribute these results to the AVS contrast project. Both sets of test patterns will contain a small readme file which describes how best to use these patterns. This data will be collected and maintained in an ongoing effort.
One word of caution in using these test patterns: It's expected that there will be considerable variation in the data collected because of unit to unit variability in specific projector models and more so between probe brands and specific models of probes. The reader should be cautioned in drawing conclusions from these comparisons in much the same way as they do in comparing ANSI contrast and on/off contrast from various sources. Relative comparisons between the same probe type and better yet the same probe type from the same forum member are likely to yield more accurate and meaningful results. If enough forum members contribute to the project it may even be possible to get a much better feel on the variability of probe brands by examining the statistics of each.