How To Interpret HardingFPA Results

 

  1. How does HardingFPA interpret the meanings of terms like 'flash' and 'stripe' that are used in the Broadcast Guidelines?
  2. The Harding FPA generates a number of graphical traces. What exactly do they all mean?
  3. Tell me more about the flash risk traces.
  4. What is the difference between a flash warning and a flash failure?
  5. The Risk Trace appeared in the warning zone (Pass with Caution) for a long sequence of images but did not enter the fail zone. Why?
  6. The Risk Trace persists for several frames after a flash occurred. Is this right?
  7. Tell me more about the Flash Diagnostic traces.
  8. The Diagnostic Trace made an upwards step where no flash was obviously observed.
  9. A flash clearly occurred but the number of Diagnostic Trace steps did not increase. Can this be right?
  10. A flash or flashes occurred (as indicated by the Diagnostic Trace) but the Risk Trace did not appear.
  11. The Diagnostic Trace step count shows 7 or more transitions but the sequence is not failing. Can this be right?
  12. The Diagnostic Trace step count decreases within a few frames of an increase and not after one moving second has passed. Can this be right?
  13. Tell me more about the spatial pattern risk trace.
  14. The Spatial Pattern Risk Trace persists at level 0.4 (warning) although the pattern is constantly changing.
  15. The Spatial Pattern Risk Trace response to a spatial pattern initially increases but persists even after the pattern stimulus has disappeared.
  16. A stationary pattern, seemingly containing bar-like patterns does not fail. Can this be right?
  17. The Spatial Pattern Risk Trace suddenly changes from Warning (Pass with Caution) to Failed with little or no apparent change in the image.
  18. The HardingFPA generates spatial warnings but it's not obvious where the pattern is.
  19. Tell me more about the Extended Analysis trace.
  20. The Luminance Flash Risk Trace or the Red Flash Risk Trace has appeared for more than 5 seconds but the Extended Failure Trace has not appeared. Can this be right?
  21. The Extended Failure Trace appeared before a full 5 seconds of Flash Risk activity is indicated. Can this be right?

 


 

 1. How does the HardingFPA interpret the meanings of terms like 'flash' and 'stripe' that are used in the Broadcast Guidelines?

These terms were OK when broadcast material was being manually checked by experts like Professor Graham Harding (a painstaking and time-consuming job); however computer-based testing requires a greater explanation of acceptable limits and precise definitions of how and when a guideline is violated.

A flash consists of 2 opposing transitions in luminance. This can either be a transition from low to high luminance followed by a transition from high to low luminance, or the other way round. Likewise, a stripe consists of 2 opposing spatial changes in luminance as you move across the stripe.

 2. The HardingFPA generates a number of graphical traces. What exactly do they all mean?

The HardingFPA Analyser graphical interface displays the results of analysis in two ways. The results of luminance flash, red flash and spatial pattern analysis indicating a level of risk are displayed as bold green, red and blue traces respectively (Risk Trace). An additional selectable option shows a diagnostic trace which identifies where a luminance transition or red transition has occurred (Diagnostic Trace). This trace is faint green or red and is stepped in profile. Each upward step represents a transition gained between two frames and contributes to the total present within one moving second. A downward step indicates that a transition has been lost from the moving second.

 3. Tell me more about the flash risk traces

The HardingFPA analysis identifies a transition of at least 20cd/m2 at pixel level after compensation for motion and noise and then searches for an opposing transition. An internal count is maintained of transitions seen by each pixel over the preceding second. Pixels which have seen several alternating transitions within the previous second, are analysed in full and it is only these which contribute to the main Risk Trace.

The flashing area must exceed 25% of the video frame to generate a flash failure; otherwise a warning (Pass with Caution) may be indicated.

 4. What is the difference between a flash warning and a flash failure?

A flash warning is indicated when a flash risk trace appears but remains below the horizontal black pass/fail line. A flash failure occurs when the flash risk trace moves above the pass/fail line.

The Risk Trace will appear when guideline criteria have been exceeded. Warnings will appear when frequency and amplitude have been exceeded and failures will appear when frequency, amplitude and area criteria are exceeded.

 5. The Risk Trace appeared in the warning zone (Pass with Caution) for a long sequence of images but did not enter the fail zone. Why?

The HardimngFPA detected flashing of above 3Hz and at least 20cd/m2 in amplitude; but the flash area is less than 25% of the screen area. If the criteria for failure are not all met then the system will generate a line based on how close the material is to failure. If the line is hovering just below the failure line then it might only need a small change in size of the flashing area to push it into failure. Such a small change could be introduced when converting between formats, frame rates or codecs if this is not done carefully.

The example below shows that the flash risk has remained close to the failure line for long enough to generate an extended flash failure. This occurs whenever more than 80% of the frames in the last 5 seconds generated flash risk warnings of 0.3 or 0.4 (i.e. close to failure).

 6. The Risk Trace persists for several frames after a flash occurred. Is this right?

This is perfectly normal, and arises because of the way the HardingFPA has to detect flashing frequencies over the most recent second. You do not need to worry about the persistence of the flash risk traces, you need to principally examine the second before the line moves into failure, or where the diagnostics trace shows that the transition count is still rising. Once you have corrected all the causes of the line first moving into failure then the material at that point will pass the test. Note however, that lots of flashing close together will generate a much longer compound failure: the best thing to do is deal with the flashes one at a time until the material passes the test.

 7. Tell me more about the Flash Diagnostic traces

The Diagnostic Trace for luminance and red is intended to provide additional feedback on the amount of flash activity present in the sequence and if the permitted number of transitions have been exceeded. A Diagnostic Trace represents the minimum number of transitions that the most active 25% of image pixels have seen over the last second. Thus a sequence analysed with 4 upward transition marks/steps (the count of 4 is also indicated at the top of the interface screen) indicates that the most active 25% of all image pixels have seen 4 or more opposing transitions over the last second i.e. 2 complete flashes. In reality some of these pixels may have seen considerably more than 4 transitions but insufficient to exceed 25% of the image.

The Diagnostic Trace is a useful guide to flash activity which can cause a violation of guidelines since all flash activity above 1.0 Hz is indicated. This information can help to select the removal of a flash or flashes to achieve compliance.

The Diagnostic Trace should be used with care since the image pixels which comprise the most active 25% are constantly changing. Thus pixels that cause the diagnostic step count to change from 4 to 5 may not be the same pixels that subsequently cause a transgression of guidelines 10 frames later. Conversely, a visible transition in an image may not necessarily lead to a higher transition count if the pixels that experience the transition are not part of the most active 25%.

 8. The Diagnostic Trace made an upwards step where no obvious flash was observed

The diagnostic trace represents the number of transitions seen by the most active 25% of the screen over the most recent second. Therefore continuous image activity (e.g. localised flashing and rapid movement within the scene caused by rapid camera pan or zoom etc) can steadily increase the number of transitions that individual pixels have seen, and when at least 25% of those have seen an extra transition will the diagnostic count increase. These changes will usually, but not necessarily, coincide with obvious transitions in the current image. Failures involving the latter can be quite tricky material to fix, and may only be possible by reducing the brightness of the image or to remove the offending sequence.

 9. A flash clearly occurred but the number of Diagnostic Trace steps did not increase. Can this be right?

Since the diagnostic trace constantly modifies (updates) the number of pixel transitions seen over the most recent second, it is possible for the HardingFPA to gain a transition from the current image and lose a transition from exactly one second previously. Accordingly you will observe a transition clearly occur in the current image but the Diagnostic Trace step count will remain the same. When this occurs, the HardingFPA inserts a squiggle (see picture with inset) to indicate that the diagnostic trace has simultaneously gained and lost a transition over the most recent second between video frames.

Other reasons why a visible transition does not increase the step count can be because the pixels that see the transition are not part of the most active 25% or because the contrast is generally less than 20&cd/m2 or because the change in luminance is part of a continuing trend rather than a reversal (as is the case in the example below).

 10. A flash or flashes occurred (as indicated by the Diagnostic Trace) but the Risk Trace did not appear

The main flash risk trace (dark green line) may not appear if the flash frequency is significantly within guideline limits. Remember that 2 opposing transitions make up a single flash. The example below shows that 2 transitions have been detected by the diagnostic trace (light green line) but that the main risk trace has not yet appeared because the flash frequency up to this point is not considered to be significant.

 11. The Diagnostic Trace step count shows 7 or more transitions but the sequence is not failing. Can this be right?

In most cases, the appearance of the Flash Risk Trace above the failure line will be accompanied by a Flash Diagnostic Trace step count of 7 or more. However, there are two exceptions to this rule. The first is when using the HardingFPA "Legacy (version 2.5)" mode as the Risk and Diagnostic Traces then operate over different time intervals. For example, in this test mode, a sequence of alternating transitions followed by a sequence of static images will cause the transition step count to remain static for several frames until discarded over the moving second. In the example below, the Diagnostic Flash step count persisted at 7 steps for several frames after the Risk Trace returned to below the failure line.

A second example is when testing content using the Japanese NAB guidelines: it is also possible for the flash risk trace to go into failure while the diagnostic trace is still in the caution zone (see image below). This can occur if one of the transitions in the most recent second is classified as a “scene change” (see bottom entry of the Advanced Information tab) where 80% of the image has seen a significant luminance transition of 20IRE units or more. When this occurs, the maximum allowable number of transitions is reduced from 6 down to 3 and, in this example, failure took place when the 4th transition was detected.

 12. The Diagnostic Trace step count decreases within a few frames of an increase and not after one moving second has passed. Can this be right?

It is true that the Diagnostic Trace monitors transitions over the most recent second but the most active pixels that determine this diagnostic count are, in most cases, constantly changing. The Diagnostic Trace will only follow this one-second pattern if the most active pixels are changing together in phase. The example below shows the diagnostic trace (light green) increasing from zero to one for a period of only two frames (around the vertical amber current frame line) before returning to zero.

 13. Tell me more about the spatial pattern risk trace.

The HardingFPA examines material for pattern structures that contain pairs of alternating light-dark bars of at least 20cd/m2 contrast. The HardingFPA will ignore chequerboard patterns but will detect arrays of repeated elements which are bar-like in structure. When a bar-like pattern is detected the HardingFPA will compute the number of pairs and the area over which the pattern exists. If the HardingFPA detects 6 or more stationary pairs which persist for more than 0.5 seconds and the area occupied is greater than 40% then a failed sequence will be indicated. If these criteria are only partially met then a warning (Pass with Caution) will be issued.

The HardingFPA will probably issue a large number of warnings due to the abundance of bar-like structure in the environment or in man-made imagery (animation etc.) Mostly these pattern structures will be insufficiently regular, may drift or contain insufficient pairs of light-dark bars for complete failure. Additionally all other guideline criteria must be fulfilled before failure occurs - area of image and pattern contrast.

Finally patterns which change direction, oscillate, flash or reverse in contrast are analysed both as a pattern and as a flashing stimulus. Such pattern stimuli might therefore indicate a Flash Risk Failure when the pattern occupies more than 25% of screen area.

 14. The Spatial Pattern Risk Trace persists at level 0.4 (warning) although the pattern is constantly changing.

A pattern must remain stationary for at least 0.5 seconds in order to fail even though all other criteria have been met. The Advanced Information tab in the example below shows that a spatial pattern has been detected with 50 cd/m2 contrast (limit 20 cd/m2) and covers 46% screen area (limit 40%). However, in this example, the system has not generated a failure because motion, caused by the camera pan and zoom, makes the detected spatial pattern exempt from failure under the Ofcom guidlines.

 15. The Spatial Pattern Risk Trace response to a spatial pattern initially increases but persists even after the pattern stimulus has disappeared

In-depth analysis is averaged over the most recent 5 frames to prevent erratic response to video noise or sudden changes in pattern area or contrast. This averaging causes the initial system response to increase with pattern duration and persist for a short while after the pattern is no longer present.

 16. A stationary pattern, seemingly containing bar-like patterns does not fail. Can this be right?

Some patterns that are clearly bar-like in one direction may also possess some local structure in the orthogonal direction. Alternatively, a pattern may not be sufficiently regular or may not have sufficient contrast throughout. Any such structure may cause the system to see fewer than 6 light-dark bars or may separate a provocative pattern into two or more regions. Either of these mechanisms can save a provocative pattern that would otherwise have failed.

The spatial pattern in the example below passes because of text and foreground objects which break up the bar-like pattern into smaller irregular regions.

 17. The Spatial Pattern Risk Trace suddenly changes from Warning (Pass with Caution) to Failed with little or no apparent change in the image.

A Spatial Pattern might not exceed all the guideline limits but may be running very close to failure. Under these circumstances, a very small change in video contrast or spatial pattern occlusion may be sufficient to cause failure. Additionally, a stationary spatial pattern is not allowed to go into failure until it has persisted for more than 0.5 seconds, in which case a transition from warning to failure can occur without any change in the video frame content.

 18 The HardingFPA generates spatial warnings but it's not obvious where the pattern is.

Occasionally the spatial trace may appear when there is no obvious spatial pattern present in the video stream. The example below shows a picture of the sea generating spatial warnings caused by waves in perspective creating faint, repeating structure. Other candidates for generating unexpected spatial responses are: landscape in perspective, net curtains and reams of paper. However, it is highly unlikely that any of these scenes would actually lead to a spatial pattern failure.

 19. Tell me more about the Extended Analysis trace

The HardingFPA continuously monitors luminance and red flash activity which is close to failure. If such flashing activity persists for more than 4 seconds, an extended failure may occur as shown below. The black Extended Failure Trace of constant height appears if such flash activity is detected in at least 80% of the most recent 5 seconds.

As shown above, this extended failure mechanism is deactivated during a full failure or if flash activity is no longer in the caution band. This means that the extended trace may appear to be erratic or intermittent if the flash activity hovers between warnings and flash failures. This mechanism ensures that the extended failure does not persist for up to a full second after the last provocative image or flash has occurred.

 20. The Luminance Flash Risk Trace or the Red Flash Risk Trace has appeared for more than 5 seconds but the Extended Failure Trace has not appeared. Can this be right?

The Extended Failure Trace reacts over a 5 second period and is activated by an accumulation of flash activity close to failure. If Flash Risk Activity exceeds guideline limits and causes an actual failure, the Extended Failure mechanism is deactivated. Likewise, if the diagnostic trace falls below the caution band (see above), the Extended Failure trace is again deactivated. The Diagnostic Trace (which incidentally can be turned on or off as required by key F5) provides useful information about the location of transitions throughout a sequence.

 21. The Extended Failure Trace appeared before a full 5 seconds of Flash Risk activity is indicated. Can this be right?

The Extended Failure trace is activated when at least 80% of the preceding 5 seconds contain luminance or red flash warnings. This could occur after as little as 4 seconds if flash warning activity is continuous.