SG9665GC High Contrast/Dynamic Range Flaw and other Discouragements

arcticfire said:
Here's a few more grabs from different sources if it helps.
Click to expand...

This is interesting stuff to look at but I don't see what it is telling us that you can't see by just looking at the imagery coming out of the camera. Since digital video tone levels are measured from 0-255, any analogue conversion that is above 255 or below zero would simply get clipped. So it is hard to understand how numbers like "IRE110" could be valid when taken from a digital camera unless it is postulating a result.
 
jokiin said:
that's a very narrow field of view on the Cobra, probably one of the narrowest I've seen used on a dashcam
Click to expand...

Yes, it is very narrow - about half the width of the SGRC or Panorama. I've not seen anything else as narrow. Even the Mobius A isn't as narrow as we might think.
The Cobra 840 is great for grabbing details of anything in its arc during daylight, but at junctions or on sharp bends vehicles can slip past outside its field of view and at night the images are fairly dark.

The Cobra 840E's detail-catching ability paired with the wide angle and good night vision of the SGRC (or the Panorama which preceded the much more compact SGRC) make a good team. The Panorama ended up as a rear cam in the family car (MPV) which suits it well. If I could only afford or fit one, it'd be the SGRC, but I am fortunate enough to be able to have more than one so I have the Cobra too for its quirkiness.
 
Please correct me if I am wrong.
For a giving dynamic range which is fixed value for a sensor. Under those light situation, the dynamic range is way out of ability of the sensor in this price range.
For a still photo application we can do a exposure bracketing with + and - EV then combine together for a real HDR photo. But for a video that's not easy doing in realtime...
Looks like GC is design to priority low light performance, so high light clip is not surprise here. But it's a dash cam, all I want is the detail on the road. So I think set AE mask base on lower half of frame is fine. For my personal opinion, sky blow out is acceptable, as long as does not do full frame blow out.
How about add a feature which overlay AE mask on the screen to help us aim the camera for better exposure.
 
you know what, seeing just a couple of pictures here, relating it to my CCTV with WDR how it works with shadows and brightness this AE table thing has just clicked into place in my head and I understand it how it works now and what jokiin is fully trying to achieve.
 
Dashmellow said:
This is interesting stuff to look at but I don't see what it is telling us that you can't see by just looking at the imagery coming out of the camera. Since digital video tone levels are measured from 0-255, any analogue conversion that is above 255 or below zero would simply get clipped. So it is hard to understand how numbers like "IRE110" could be valid when taken from a digital camera unless it is postulating a result.
Click to expand...


I really don't know enough about the technical side of this other than what I've briefly read earlier online but I did read something where it mentioned that sometimes cameras are deliberately set above 100 to 110 for creative purposes to deliberately overexpose certain things.

From this and my lack of understanding I can only approach it from a perhaps logical assumption that in digital it has either been done in error to try and increase the dynamic range somehow or force a shift in the dynamic range (although that doesn't really make sense to me) or that it's a setting or trait of the camera that has been missed by the manufacturer? Or maybe the logic by the manufacturer with this was sound but the hardware simply can't cope with it?

I'm really not sure but I feel it's something that may be worth eliminating or fixing that may help. The only thing I'm fairly sure of is that I don't feel that it's necessary or good for the camera to allow the exposure to go to this level and it could be worth looking to see if it helps fix things?

I'm probably just grabbing at straws but from a logical point of view to me with what I've seen so far it would seem to support or perhaps explain your theory that the tonal curve is going wrong in this area? I'm interested to see what the manufacturers thoughts are about this 110 level and why the camera is allowed to get there.

I'm tempted to try and learn more about it but I've so little time and to be honest any conclusions I come to would still be speculation so probably not much point and best to let the manufacturer look at it as maybe they'll have the ability to adjust it and actually see the immediate results. :)
 
painkiller said:
But it's a dash cam, all I want is the detail on the road. So I think set AE mask base on lower half of frame is fine.
Click to expand...

agree but the current setting means there is far too much emphasis placed on the lower section, we need to fix that

painkiller said:
How about add a feature which overlay AE mask on the screen to help us aim the camera for better exposure.
Click to expand...

don't think we could do that but once it's sorted we would be able to give you a better idea of how the aim will affect things
 
Interesting article here: http://www.kenstone.net/fcp_homepage/video_levels_nattress.html

In particular this bit:

In the ITU-R BT.601-4 specification, reference black is at code 16 and reference white is at code 235. Codes 0 and 255 are “reserved” and not used for video encoding. Codes 1 though 15 are for “super black” and codes 236 through 254 are for “super white”. The super black and super white codes are footroom and headroom for the accurate preservation of transient peak excursions beyond the legal video range. It should be noted that many digital video cameras routinely generate super white (above white, or above code 235) video levels and these need to be dealt with before such video can be safely broadcast.
Click to expand...
 
all this stuff jokiin is learning about dashcams will only assist in development of future out of this world streetguardian cams that will keep trumping the best out there to make streetguardian remain one of the tops brands available.
 
jokiin said:
agree but the current setting means there is far too much emphasis placed on the lower section, we need to fix that
don't think we could do that but once it's sorted we would be able to give you a better idea of how the aim will affect things
Click to expand...

Thanks for that. I thought we just need add a horizontal line in GUI overlay. So SDK may not allow it then...
 
wozzzzza said:
all this stuff jokiin is learning about dashcams will only assist in development of future out of this world streetguardian cams that will keep trumping the best out there to make streetguardian remain one of the tops brands available.
Click to expand...

anytime you think you're the smartest person in the room, you're in the wrong room, there is always something to learn, there's some really impressive tech coming that will find its way into products over the next 12 months or so
 
arcticfire said:
Interesting article here: http://www.kenstone.net/fcp_homepage/video_levels_nattress.html
Click to expand...

In particular this bit:
In the ITU-R BT.601-4 specification, reference black is at code 16 and reference white is at code 235. Codes 0 and 255 are “reserved” and not used for video encoding. Codes 1 though 15 are for “super black” and codes 236 through 254 are for “super white”. The super black and super white codes are footroom and headroom for the accurate preservation of transient peak excursions beyond the legal video range. It should be noted that many digital video cameras routinely generate super white (above white, or above code 235) video levels and these need to be dealt with before such video can be safely broadcast.
Click to expand...
Your quote is from the same article I quoted from earlier regarding the difference between analogue and digital video signals. It is important to consider that this quote you cite is discussing broadcast video in PAL and NTSC as rendered in Final Cut Pro editing software. The very next paragraph in the article discusses how "the computer monitor is very different to a video monitor (PAL or NTSC) in many ways. Computer monitors often have a different “gamma” (gamma is the name given to the non-linearity in the values that represent the video image) to video monitors, and they handle video levels differently."

"Super white" often refers to specular highlights that contain no image detail but are part of almost any sunlit image.

Quoting text out of context from an article on editing video for braod cast doesn't really help our purpose here and as interesting as this subject may be, IRE does not enlighten us in any way that has much to do with the actual problem being discussed here, which is digital video as displayed on standard video monitors in the context of a problem inherent in the SG9665GC sensor/DSP platform.

Those green graphs you posted do look cool though. :)
 
I'm thinking that if AE table believes IRE110 to be the clipping point of highlights then logically it could think or be coded to think that IRE100 - IRE109 is still in the unblown region and acceptably exposed and subsequently effect the rest of the exposure also as it will all be sitting slightly wrong hence why SG footage does always seem to be slightly more bright? If it were a monitor issue then wouldn't I be able to pull some of the highlights back? Or even the graphs would show that the exposure is sitting correctly at IRE100 and under even though my monitor may not be able to display it correctly?

It just seems a little coincidental that the mini doesn't do it but I think it would be interesting to get some raw footage from other cams to see what they do?
 
jokiin said:
that's a very narrow field of view on the Cobra, probably one of the narrowest I've seen used on a dashcam
Click to expand...

Although the Cobra outperforms almost anything else for detail capture in daylight, it's well below the SG/Panorama at night.

First screenshot is the Cobra. See the red car? No, neither can I.
Second screenshot is Panorama (image very similar to SGRC). Aha! Indeed it's a red car!
-



 
arcticfire said:
I'm thinking that if AE table believes IRE110 to be the clipping point of highlights then logically it could think or be coded to think that IRE100 - IRE109 is still in the unblown region and acceptably exposed and subsequently effect the rest of the exposure also as it will all be sitting slightly wrong hence why SG footage does always seem to be slightly more bright? If it were a monitor issue then wouldn't I be able to pull some of the highlights back? Or even the graphs would show that the exposure is sitting correctly at IRE100 and under even though my monitor may not be able to display it correctly?

It just seems a little coincidental that the mini doesn't do it but I think it would be interesting to get some raw footage from other cams to see what they do?
Click to expand...

Analogue IRE has NOTHING to do with digital video. No competent firmware engineer would bother with it or make such voltage conversions when they are working on 8 bit digital AE tables.
 
Last edited:
I'm only using IRE as reference instead of saying 235 and 255 etc.
 
I had an idea: Could it be that the reflections from the bonnet are the cause of the whiteout on my cam?
MrQUnBA.jpg
 
No it should be the opposite actually. The metering should see that the scene is too bright and adjust the exposure accordingly.
 
You must log in or register to reply here.
Back
Top