Feitelijk's SG9665GC

[Dashmellow]

No, unless you consider an overcast day to be yellow.

Here's a full 5 minute clip that has a few instances. At the beginning it's already in a 'blue shift' and normalizes in a few seconds. Between about 1:15 and 2:00 you can see 2 more occurrences. At about 2:30 the image has a less severe blue shift that's really noticeable as I complete the left turn at the bottom of the exit ramp and it normalizes - that one could be impacted by the exposure changing pretty dramatically at the same time.

If you watch the pavement color you can see the white balance continually adjusting.

Specifics:
F/W: V3 Beta 30
A/E: 0, Center Weighted
Contrast: -2
CPL

The color shifts are relatively subtle but I definitely see them. Usually when this happens in dash cam footage the camera has encountered a large object or lighting situation that is the exact opposite RGB value of the color shift on a standard RGB color chart as I talked about in THIS post about a month ago where I showed this on the RGB charts and discussed the phenomenon. That's why I asked. In this case, with the exception of a small yellow road sign near the 1:15 mark and a general yellowish warm tone to the pavement in your video there is nothing obvious to point to that that camera might be reacting to. Not sure of what the explanation might be but so far I have not flashed V30 and think I will hold off for now as I've had a fairly positive experience with beta 27 considering the dynamic range/high contrast issues I've been concerned with for so long.

As I mentioned in that post the Mobius had an issue with earlier firmware where driving past a Red building or large Red vehicle would cause a shift to Green/Cyan (Opposite of Red on the RGB chart) and then return to normal as soon as the Red object was out of view. There was a similar Blue/Yellow shift issue as well.

 

[Feitelijk]

Lover EV seems to help a bit, but still cant read them.

 

[Feitelijk]

This is a daytime comparison with the G90C (on 1080p) (SG9665GC on -1EV and -1contrast)
The GC is less noisy

 

[DT MI]

The color shifts are relatively subtle but I definitely see them.....

Actually they're not all that subtle when you speed up the video. Here's the same clip at 10x speed.

 

[Dashmellow]

Actually they're not all that subtle when you speed up the video. Here's the same clip at 10x speed.

Appears to be a private video.

Just so you know, I used the word subtle not because you can't see the color shifts (as I acknowledged) but because the shifts in your video are much less pronounced than in many other examples of these kind of color shifts that I've seen or experienced personally.

 

[DT MI]

Appears to be a private video.....

Meant for it to be unlisted. Fixed now.

...Just so you know, I used the word subtle not because you can't see the color shifts (as I acknowledged) but because the shifts in your video are much less pronounced than in many other examples of these kind of color shifts that I've seen or experienced personally.

I assumed that's what you intended, it's just a bit more dramatic when the video is accelerated.

This entire subject of color shifting intrigued since I've just recently become aware of it in my cameras so I've been looking back at some of my videos. In summary this is what I've found so far (in no particular order and for whatever it 's worth):

1) Shifting occurs with both V1 and V2 cameras
2) Blue/Yellow (and to a lessor degree Green/Red) shift has been happening as long as I've had a GC (oldest files I have go back to June, '15).
3) Blue/Yellow shift seems to happen more in overcast conditions
4) Many times when the car is started outdoors when overcast the video has a distinct blue cast from the first frame
5) I've only seen a Green/Red shift in sunshine
6) Green/Red shift is less pronounced and more infrequent than Blue/Yellow
7) The only time I've seen the camera go into 'Smurf Mode' (all blue) is since upgrading the V2 to Beta 30 firmware
8) I've seen no evidence of shifting at night
9) I've seen no shifting when the weather is overcast and most cars have headlights on

 

[jokiin]

This is a daytime comparison

when you create the gifs could you use a longer delay, the rapid switching between the pics makes it harder to compare

 

[Dashmellow]

View attachment 28777

Meant for it to be unlisted. Fixed now.


I assumed that's what you intended, it's just a bit more dramatic when the video is accelerated.

This entire subject of color shifting intrigued since I've just recently become aware of it in my cameras so I've been looking back at some of my videos. In summary this is what I've found so far (in no particular order and for whatever it 's worth):

1) Shifting occurs with both V1 and V2 cameras
2) Blue/Yellow (and to a lessor degree Green/Red) shift has been happening as long as I've had a GC (oldest files I have go back to June, '15).
3) Blue/Yellow shift seems to happen more in overcast conditions
4) Many times when the car is started outdoors when overcast the video has a distinct blue cast from the first frame
5) I've only seen a Green/Red shift in sunshine
6) Green/Red shift is less pronounced and more infrequent than Blue/Yellow
7) The only time I've seen the camera go into 'Smurf Mode' (all blue) is since upgrading the V2 to Beta 30 firmware
8) I've seen no evidence of shifting at night
9) I've seen no shifting when the weather is overcast and most cars have headlights on

Thanks for fixing the video!

Interesting....You're right, I don't think I've ever seen any color shifting at night although it may be harder to tell. I've not experienced any color shifting with the GC but I am still on beta 27 until such time as I have the opportunity to do some thorough testing. Until then I'm sticking with beta 27 because it's given the best results I've had with this camera.

As I've been trying to show, in virtually every instance of dash camera color shifting I have ever seen personally on my own cameras, here on DCT and on YouTube the camera encounters an object or scene that is the exact opposite color on the standard RGB color wheel and then seemingly overcompensates in the other direction past the required amount of color correction until opposite colored object or scene leaves the image.

 

[Feitelijk]

when you create the gifs could you use a longer delay, the rapid switching between the pics makes it harder to compare

Ok. I found also some gif artefacts like extra noise and stray pixels I can't get rid of. perhaps this is not the best way to compare.

 

[Feitelijk]

Tried at night on EV -2 and contrast -1
Seems a bit better. but quite dark obviously.

 

[Feitelijk]

Oh yeah.

I wanted to know about the motionblur, but as you can't say anything about that without knowing the motion of the plate I decided to measure that.
After comparing screenshots for two whole minutes I got bored and decided I needed a licenceplatetracker.
And here we are:

Original: https://drive.google.com/open?id=0BxJHQyAct4u4Wjd2NFJBTUhlUzg

First results, in the dark motionblur makes the plateillegible when there is a shift of more than 5px/frame

 

[Feitelijk]

EV -1

EV -2

 

[Feitelijk]

When the motion is low enough the plate is too far away to be legible, when it gets close enough the speed is too high.

 

[Feitelijk]

And a similar movie on contrast -1 instead of -2

 

[Feitelijk]

Comparing the two contrast at the same speed

 

[Feitelijk]

Still a problem. (EV -1, AE center , contrast -1, others default)

Recovery 5 minutes and 8km further.

 

[Feitelijk]

Low cut is not optimal in the other car.
ae=lowcut cont=-1 ev=-1

 

[Feitelijk]

And this is how much Venus moved in one week:
AE low EV 0
vs
AE center EV -1

 

[Feitelijk]

White bright clouds are still troublessome on the lowcut (EV0, cont -1)

 

[Feitelijk]

Sodiumlights:

But the G90 is quickly recovered.