The best AE metering pattern for the most dashcams.

nutsey

Well-Known Member
Joined
May 12, 2015
Messages
627
Reaction score
574
Location
Turkey
Country
Russian Federation
Does it exist?

From where I stand it's something like this for hypothetical 4x4:

0 1 1 0
1 2 2 1
2 3 3 2
0 0 0 0

What do you think? How should it look for 8x8 or 12x8?
 
I would swap the top and bottom lines.

Of course it depends on your vehicle and what you intend to use the video for...
 
I see it like this:

0 0 0 0
3 2 2 3
3 2 2 3
1 0 0 1

During night there is too much light at low-central 2 2 zone because of car headlights.

enjoy,
Mtz
 
Last edited:
I would swap the top and bottom lines.
Does it mean that the proper hood exposure is more important than details in the sky?

During night there is too much light at low-central 2 2 zone because of car headlights.
And what about those 3's on the side of central 2 2? I guess these zones are even more sensitive to the headlights during night time.
 
In most of the night video recordings I see much brightness on the center and dark on the sides.
The sky is 0 important for me and even more, because of the too much brightness added to the firmwares the sky is over exposed. Also the car license plates are overexposed.

I am new in this world of exposure tables, I have no experience and right now these are my suppositions.
If I would have a firmware with different exposure tables for Viofo A119S I could make a side by side comparison.

enjoy,
Mtz
 
  • Like
Reactions: gse
Does it mean that the proper hood exposure is more important than details in the sky?
I don't have the hood in my image, and plates are more important than the sun or clouds, normally, unless I am making a video rather than capturing evidence:

4.jpg

And what about those 3's on the side of central 2 2? I guess these zones are even more sensitive to the headlights during night time.
Depends on the situation, in this situation I want those threes to be zeros because my headlights don't illuminate out there so the threes make the centre part overexposed, and I might want something in the sky to get a nice exposure on the moon and venus:

3.jpg
 
Last edited:
I don't have the hood in my image
I like your approach as it allows to get as many traffic lights as possible without losing any important details on the road itself. But many dashcam manufacturers recommend sky to road ratio of 30% to 70% and they use AE-metering tables to match it.

Look at the passage of the tunnel.
What AE pattern was used for the 'Alfsoft'?
 
What AE pattern was used for the 'Alfsoft'?
Alfsoft
0001020202020100
0101020303020101
0101010404010101
0102020505020201
0103040404040301
0001020202020100
0000000000000000
0000000000000000

I moved to the point, I like it more !!
Point
0000000000000000
0000000000000000
0000000101000000
0000010303010000
0000010303010000
0000000101000000
0000000000000000
0000000000000000

There also depends on the tables of excerpts, not all DVR is suitable.
 
But many dashcam manufacturers recommend sky to road ratio of 30% to 70% and they use AE-metering tables to match it.
70 vs 30 is just too much, in reality nobody will set the dashcam like this. Some of them which want more road will use maybe 60-40, but most of the people which knows how to mount a dashcam are using 50-50 or 55-45. I consisder this 70-30 is an old story from poor CMOS and poor manufacturers which could not setup a metering table.

Look at the passage of the tunnel.
Nice comparisons but a firmware should not be optimized based on tunnels which are very rare in normal driving. Firmware should be setup for normal situations and if in a tunnel situation the video is not OK we should accept like it is, but not to ruin a firmware which will be used 99% in normal situations with poor results and 1% in tunnels with good results. I would sacrifice the tunnel but to have good image in other situations.

@vvs49 Are you modifying Novatek firmwares in that video comparisons with tunnel?

enjoy,
Mtz
 
in reality nobody will set the dashcam like this
I will. But only for some certain research reasons. And I don't believe that any one-cam setup is able to do the job here (note the modern smartphone market).
 
  • Like
Reactions: Mtz
I don't have the hood in my image, and plates are more important than the sun or clouds, normally, unless I am making a video rather than capturing evidence:

View attachment 37638


Depends on the situation, in this situation I want those threes to be zeros because my headlights don't illuminate out there so the threes make the centre part overexposed, and I might want something in the sky to get a nice exposure on the moon and venus:

View attachment 37639

Daylight shot is great. Nightime, less so. See you got your hands on A129 Nige.
 
Are you modifying Novatek firmwares in that video comparisons with tunnel?
No, I put Novatek NT96658 for testing and I do not regret it yet.
But I turn to the DVR with NT96663 and soon will be testing, it's more difficult.
I will change the exposure.
Table Point - from V747W
 
Daylight shot is great. Nightime, less so. See you got your hands on A129 Nige.
It was dark in that night time frame! What is it that you don't like about it?

Compared to the B1W (first image), it is both brighter and a faster shutter speed:
5.jpg
6.jpg
 
It's brighter but the 1st image is slightly sharper in the hedgerow by the tree and the road. That makes me think any vehicle will probably be the same as well. Also, a difference in colour balance - in the top one it looks like you have HID, the bottom one halogen. Were they different cars or side by side mounted in the same car? That said, it's still good in zero light, there doesn't appear to be any noise as you said and it's definitely brighter and so possibly a more sensitive. You're second capture looks less noisy that the one in your first post.
 
It's brighter but the 1st image is slightly sharper in the hedgerow by the tree and the road. That makes me think any vehicle will probably be the same as well. Also, a difference in colour balance - in the top one it looks like you have HID, the bottom one halogen. Were they different cars or side by side mounted in the same car? That said, it's still good in zero light, there doesn't appear to be any noise as you said and it's definitely brighter and so possibly a more sensitive. You're second capture looks less noisy that the one in your first post.
  • The image in the first post was posted to show the lack of headlight illumination at the sides, which the B1W image doesn't show!
  • All images are the same car, same time to within a frame or 2.
  • B1W shows the time an hour out because it did an automatic daylight savings time correction even though I didn't ask it to, A129 is on GMT like most of my cameras.
  • A129 has GPS, B1W doesn't so it's surprising that the recorded times are within a second.
  • B1W is mounted on the right of the windscreen behind the sun visor, the A129 is under the mirror.
  • Headlights are standard halogen bulbs with 13 years of use.
  • A129 has a CPL fitted, B1W doesn't.
  • The A129 image in the 2nd post is the same image as in the first post, but with the brightness turned down in a photo editor!
  • A129 is a prototype, and the images may not reflect the final product.
  • B1W I think is from the first production run.
 
Nice comparisons but a firmware should not be optimized based on tunnels which are very rare in normal driving
Deed it's not in the tunnel, but in the light - from the shadows to the bright light.
The street is in a deep shadow, and the intersection is flooded with sun - often.
I often see how the overexposure cleans up the details.
At such table while all is OK.
 
From your video, in first 6 seconds the bottom left was best.
From the middle of the video the best is top left but seconded close by the bottom left. Not so easy to observe this second part of your test because the videos are not exactly at the same frame.

enjoy,
Mtz
 
From your video, in first 6 seconds the bottom left was best.
From the middle of the video the best is top left but seconded close by the bottom left. Not so easy to observe this second part of your test because the videos are not exactly at the same frame.

enjoy,
Mtz

I agree. Watching the video, rather than observing the posted stills, going into the first tunnel, the bottom left was better for detail. In the tunnel Alsoft was best. Emerging from the Tunnel, bottom left was best again as it was the least overblown and fastest to react, in the daylight after emergence, top right looks the best to me for bright daylight - seems better exposed - 50% between Alsoft and bottom left, and seems sharper with more visible detail eg in the bush. Maybe the ideal situation is more than 1 metering pattern the camera can switch between according to the lighting conditions.

With the second tunnel, Alsoft again wins for me in the tunnel. Bottom left emerging, and this time Alsoft when just out.

My conclusion from this would be Alsoft is the best compromise for night between detail and lighting. Bottom left best maybe for high contrast situations (?) based on the transition between tunnel and light (be interesting to see how this translates into dusk situations or if the win is restricted to tunnel vision!), and in bright sun, it's between Alsoft and top right, which curiously come out with 1 best picture each!

One final observation though, based on the emergence from the 2nd tunnel, the bit rate could possibly do to be higher - some of the trees / grass are quite smudged which I'm interpreting as possibly compression.
 
Back
Top