What do "filter width" and "filter height&quo
Posted: Thu Dec 01, 2005 2:29 am
I am a newbie, and so far I can only seem to get Modect to operate in one of two modes - either the camera stays in a permanent state of alarm, or it stays in a permanet state of idle. I clearly do not understand what all the knobs really do...
I am experimenting with a single DCS-900 between 6 and 7 fps, which either because of the lighting conditions or because it's not a very good camera seems to create a rather noisy picture.
In describing filter width and height, the doc says the feature "removes any pixels that do not participate in a contiguous block of pixels above a certain size." I do not understand what this sentence is trying to say - I can think of two possible meanings, but I haven't been able to figure out which (if either) is correct.
Both of the meanings require that a pixel have 8 "contiguous" neighbors (up/down/left/right/4 diagonals) rather than 4 (up/down/left/right)...
Suppose I have a 3x3 filter. The doc could mean
1) Look at the 3x3 block of pixels centered on the alarmed pixel. If there are no other alarmed pixels in the block, ignore the alarmed pixel.
This would account for the recommendation that filter sizes be kept small, since the larger the block the more likely it will be to contain alarmed pixels created by noise, so larger blocks wouldn't tend to eliminate noise, they'd tend to retain it. But it would also imply a 1x1 filter should eliminate all alarmed pixels, which doesn't seem to be the case.
2) If the alarmed pixel is not part of a block of alarmed pixels that is at least 3x3 in size, ignore the alarmed pixel.
This would also account for the recommendation that filter sizes be kept small, since the larger the filter the more likely "real" alarmed pixels would be to be tossed. And the maximum filter size allowed (15x15) seems to greatly eliminate noise-created alarms.
I am experimenting with a single DCS-900 between 6 and 7 fps, which either because of the lighting conditions or because it's not a very good camera seems to create a rather noisy picture.
In describing filter width and height, the doc says the feature "removes any pixels that do not participate in a contiguous block of pixels above a certain size." I do not understand what this sentence is trying to say - I can think of two possible meanings, but I haven't been able to figure out which (if either) is correct.
Both of the meanings require that a pixel have 8 "contiguous" neighbors (up/down/left/right/4 diagonals) rather than 4 (up/down/left/right)...
Suppose I have a 3x3 filter. The doc could mean
1) Look at the 3x3 block of pixels centered on the alarmed pixel. If there are no other alarmed pixels in the block, ignore the alarmed pixel.
This would account for the recommendation that filter sizes be kept small, since the larger the block the more likely it will be to contain alarmed pixels created by noise, so larger blocks wouldn't tend to eliminate noise, they'd tend to retain it. But it would also imply a 1x1 filter should eliminate all alarmed pixels, which doesn't seem to be the case.
2) If the alarmed pixel is not part of a block of alarmed pixels that is at least 3x3 in size, ignore the alarmed pixel.
This would also account for the recommendation that filter sizes be kept small, since the larger the filter the more likely "real" alarmed pixels would be to be tossed. And the maximum filter size allowed (15x15) seems to greatly eliminate noise-created alarms.