The night vision potential is fantastic! The reality is not there yet.

[country_hick]

I made a very simple and easy to make mistake. I forgot to take off the mobius maxi lens b cover. I recorded a large amount of (extremely boring) footage showing the quality of the inside of my lens cap. It finally occurred to me that the lens cap was on. While going down the road (coincidentally at twilight the absolute best time for this demonstration) I removed the lens cap. I was shocked by the result.

When I removed the lens cap I thought nothing of it. After looking at the footage i discovered that with some firmware tweaks the night footage could be fantastic. For a short period of time of under 3 seconds going from complete lack of light to twilight the maxi showed great detail and color. Then once the sensor adapted to the low light conditions the detail and color disappeared.

The first 4 seconds of video show the quality of the inside of the lens cap. After that you can see the serious light gathering potential for about 1 second before the firmware settings take the details and color away.

 

[country_hick]

The video at 4.1 seconds (taken at at twilight) shows this scene.



Compare it to this screenshot taken approximately 1 second later.


Just imagine what firmware tweaking could do to improve what we could see in low light videos!!!

 

[c4rc4m]

I think what you're seeing is the difference in shutter speed and electronic gain.

In the 1st picture, there's quite a bit of blur and some grain, and given the lens cap was on, it's likely the camera had set a very slow shutter speed and maximum electronic gain, as with the lens cap on it was metering total darkness. When the lens cap was removed, it detected more light and so adjusted itself down to a faster shutter speed and less electronic gain, for maximum quality, hence the 2nd picture being sharper and less grainy than the 1st.

The 1st picture was just captured in the period between the camera being set for total darkness and twilight, with a slow adjustment reaction being the cause, hence why it disappeared in the 2nd picture.

Although the 1st picture captured on the "wrong" settings shows better low light performance, it is at the expense of the sharpness and graininess of the picture.

 

[DashcamDPR]

not to mention in the first photo, the camera is doing absolutely nothing for the glare caused by the headlights, streetlights and reflected light, so even if the cam allowed a constant much lighter shot at this time, you'd have massive glare throughout the journey

 

[TonyM]

One of the things I like about the Maxi is that it preserves highlight detail in such situations, whereas other cameras tend to overexpose the bright areas. That also means that it maintains a shorter exposure time, resulting in less motion blur.

Having said that, I think there is some scope to increase the night time exposure a little bit and open up the shadow areas without having too much impact on the highlight detail and motion blur.

 

[SawMaster]

Glare, details, and image crispness do matter, but they become less relevant when things go so dark that you lose the bigger picture. Everything is a trade-off, but it does seem that there is a great potential for low-light performance here if someone can only dig it up and make it work that way. This would probably be found in the programming and not the user settings. Too bad that isn't user-accessible the way settings are.

Phil

 

[country_hick]

There is a variety of potential.

This shot shows the oncoming headlights with less glare than existed initially, but it may still be to much.




This shot shows just a little less glare while still maintaining decent color and detail.

 

[SawMaster]

There's a perfect balance between brightness and image clarity in there somewhere. Of course what it exactly is will vary per user and what they prefer.

The front cam of my workvan serves as something of a security cam here. At night, that old Mobius can only 'see' a person's shape, but my G1W/S could see some details because the image is brighter, even though overall it has worse image clarity and a far less crispness of the image. Mobius = "I see a man of average height and build" while G1W/= "I see a man of average height and build wearing a light colored shirt and what seems to be blue jeans". If you needed to find that man, which description would you prefer to have?

Thus my position that you need enough brightness first before the other things begin to matter. What you can't see can't help you.

Phil

 

[TonyM]

Out of interest, I have run the dark and bright frames from the original post through a photo editor (Lightroom).

The bright frame has clipped quite a lot of the highlights (shown in red) and only a little shadow clipping in the bottom right corner (shown in blue). The histogram illustrates this with highlights peaking at the far right of the graph, a fairly flat profile and a small shadow peak at the far left.


By contrast, the dark frame has negligible highlight clipping on the headlights, but significantly more loss of shadow detail around the lower edges.


The final frame posted by @country_hick yesterday has better balance overall with greater retention of shadow detail.


This retained detail can be shown by increasing exposure only in the shadow areas:

 

[country_hick]

Hopefully the developer will see this thread and will be able to enhance the firmware settings to get better results that should be similar to the last frame I posted. I went through the video until I found what I thought looked like the best balance of glare from the headlights while also capturing the most details.

I have felt that the maxi's pictures in dark surroundings have been to dark. Maybe with this information the maxi can develop better night vision without making unnecessary compromises in image gathering.

 

[c4rc4m]

Out of interest, I have run the dark and bright frames from the original post through a photo editor (Lightroom).

This retained detail can be shown by increasing exposure only in the shadow areas:

Whether such selective gain can be applied in the firmware or is limited to photo editing software only is the question. I'm sure Jokin may be able to answer this. I've always understood the gain was for the whole sensor and not selective. As for contrast / brightness, I'm not aware it can be selectively applied in a camera environment, especially something as comparatively un-powerful as a dashcam. Remember with photo editing software you have a whole pc behind it and even then you're applying it to a still of a single frame. With in camera video, you're going to have to identify the area of darkness which will be constantly changing in each and every frame, at a rate between 30-60 times per second (fps) and then make the adjustments to the output side of every single frame on the fly also at the same rate. I doubt it's possible, hence the more global settings usually seen. However, I know little of the firmware and could be entirely wrong here.

 

[dirkzelf]

The video at 4.1 seconds (taken at at twilight) shows this scene.

View attachment 41539

Compare it to this screenshot taken approximately 1 second later.

View attachment 41540
Just imagine what firmware tweaking could do to improve what we could see in low light videos!!!

I've experienced the same thing with the Xiaomi Mijia dashcam

https://dashcamtalk.com/forum/threa...-imx323-gearbest-com.28266/page-3#post-386126

... Very frustrating, no one there is listening. Great potential hardware, but they seem to deliberately scr@w it up. Totally weird.

 

[Rajagra]

Even the darkest of the images can be adjusted to bring out detail. Here I've allowed the headlights to burn out, and then "equalised" the brightness.

Perhaps pointing your camera down a little would give the best overall exposure. No point exposing for the sky in the evening!

 

[c4rc4m]

As I said in my post above, it might not possible. We really need feedback from a manufacturer.

The issue is what can be done in post in a photo editor with the power of a full pc and infinite time to process the change, is potentially very different to what can be done live on the fly in the dashcam. As I said above, I believe that the application of selective gain and / or brightness control might be very difficult if not impossible on the current hardware. I don't profess to know the answer, but rather it's my gut feeling based on what I do know. I rather suspect the only current application is universal application which then blow out or increase grain in other areas.

 

[Nigel]

As I said in my post above, it might not possible. We really need feedback from a manufacturer.

The issue is what can be done in post in a photo editor with the power of a full pc and infinite time to process the change, is potentially very different to what can be done live on the fly in the dashcam. As I said above, I believe that the application of selective gain and / or brightness control might be very difficult if not impossible on the current hardware. I don't profess to know the answer, but rather it's my gut feeling based on what I do know. I rather suspect the only current application is universal application which then blow out or increase grain in other areas.

It's called WDR...

 

[c4rc4m]

That helps but how many sensors are really WDR and how many simply are software exposure compensated with a higher contrast / brightness setting to appear to be WDR and blow highlights or appear too dark in areas as a result? The fact that many softwares have a WDR setting makes me suspect many are pseudo WDR.

There are decent ways of achieving WDR with non WDR sensors such as by combining several different exposures, however, I've never heard of this for video only stills and with longer exposures, blur may become an issue because of the movement involved in dashcam recordings.

 

[Rajagra]

As I said in my post above, it might not possible. We really need feedback from a manufacturer.

The 'equalize' function I mentioned is very simple mathematically, certainly compared to compression that is already being done. It doesn't give attractive results in many cases, but it does allocate fair amounts of information to areas of differing brightness, which is what I personally want from a dashcam - information, not cinematography.

I think all of these adjustments are so closely tied to the compression algorithm that they essentially become the same thing, we just need to decide what we really want and let the manufacturers know.

 

[Nigel]

That helps but how many sensors are really WDR and how many simply are software exposure compensated with a higher contrast / brightness setting to appear to be WDR and blow highlights or appear too dark in areas as a result? The fact that many softwares have a WDR setting makes me suspect many are pseudo WDR.

There are decent ways of achieving WDR with non WDR sensors such as by combining several different exposures, however, I've never heard of this for video only stills and with longer exposures, blur may become an issue because of the movement involved in dashcam recordings.

It is HDR that uses multiple exposures, WDR can be done with any sensor, it is just a non-linear mapping of the 10 bit sensor data into 8 bit video data.

 

[jokiin]

HDR is great, for still photography

 

[kamkar]

Indeed. though having forgotten HDR was ON and you go creative, you can get some really messed up results.