PPF vs Lighting VariancesBy John Honovich, Published on Oct 08, 2012
A major flaw in using Pixels Per Foot (PPF) to specify surveillance cameras is how easily the metric is undermined. Theoretically, you should specify a PPF number, say 50, and know that you will be able to capture facial details regardless of the camera or situation.
The great hope of PPF is that it is an independent, cross industry metric. Unfortunately, it is not. In this note, we show a practical example where even modest lighting variances undermine the accuracy of the metric.
The image below is from a 10MP camera that a manufacturer uses in their standard marketing. We have zoomed in and overlayed the image detail for two portions of the scene (A and B) to demonstrate differences in actual image quality. Take a look:
Try it yourself with the original 10MP image. Digitally zoom and pan to compare the image quality of different portions of the scene.
Notice how the facial details in zone A are much more detailed than those in zone B, despite zone A having less PPFs than zone B.
Cameras are sensitive to both the amount of light in the scene and the variance of light within the Field of View. While the image as a whole is in fairly ideal conditions - not low light nor faced with direct sunlight - zone A is evenly, brightly lit, while zone B, underneath the awning is darker.
Though some cameras are better than others in WDR, this common situation, will regulary cause variances in image quality, regardless of a PPF specification.
Hurts Megapixel More - Indirectly
While any camera faces lighting variances, megapixel (especially multi-megapixel) ones are more likely to find this to be a practical problem. Since megapixel cameras often are used to coverage larger areas, whether it be through fisheye or simply wide angle lenses, the larger the area covered, the more lighting variances are likely. The example above shows that even in a hand picked manufacturer marketing sample, such variances can still impact details captured notably.
What to Do?
One either needs to overspecify the PPF level (i.e., 70 instead of 50, etc.) or accept that some areas of a scene with fail to capture the level of detail projected due to lighting variances. Ultimately, this is not just a specific camera problem but a fundamental flaw of the PPF methodology (though cameras that are better at WDR or night time imaging can somewhat help minimize these problems).
This, of course, gets far worse when dealing with very demanding situations, like low light or direct sunlight. See our PPF test results for more.
Lessons from IPVM Class
This issue was raised in an excellent discussion during a class of the IPVM Advanced Surveillance Course (hat tip Rod for calling out the specific example above).
Earlier in the course, we asked the following question:
The right answer is that you cannot abstractly know how many pixels per foot you need to capture a face until you understand the scene conditions. The example above is an excellent example of that. You might just 'throw more pixels at it' by specifying a higher metric (say 60ppf instead of 40ppf) but it's still an imprecise (and potentially inefficient) way of doing so.
Unfortunately, this shows that there is no easy way out if you want to both capture high quality details and not 'waste' pixels. Carefully evaluating site conditions and setting the right customer expectations is essential.