Super Low Lux Geovision Minidome Tested

Super Low Lux Minidome Tested

Minidome, or microdome....?

Hi Derek, thank you for testing this very interesting camera. It claims to have WDR. Did you test it's WDR capabilities? Which length of lens did you use for testing? Thank you for your help.

Am I crazy or does it seem that on the graphic for factory focus vs. ipvm focus, that the effective aspect ratio significantly changed also? Is that really possible with a symmetric lens?

This is great, thank you!

Lens F number and image sensor (pixel) size are the most important for estimation of light sensitivity.

Just compare images on 2 lux with these parameters:
GV MFD1501 F1.5 1/3"
AXIS M 3004 F2.8 1.4"
Bosh Flexidome F1.2-1.6 1/3"
Dahua HD2100N 1/3 F1.8
Hickvision HS-2CD7164-E 1/3 F1.8

Bosh Flexidome has shown result worse than expected. Other cameras shown results according to their F-number and sensor sizes.

In the Conference Room (<1 lx) GV MFD1501 turns on hight AGC+noise reduction therefore its image seems more bright than HD2100N and DS-2CD7164-E.
In other respects sensitivity of GV MFD1501, HD2100N and DS-2CD7164-E is the same. If HD2100N and DS-2CD7164-E turn on the Hight AGC+NR, their images must be simular to GV MFD1501.

Before performing any image processing we need to get initial image information. In low light the number of photons caught by each pixel is directly proportional to square of this pixel, the size is proportional to image sensor size. At the same time noise of the pixel is not directly proportional to the size. Therefore the more the size is the better is signal/noise ratio of initial image.

Influence of the pixel size is simular to influence of lens F stop. The more F stop is the less photons are caught by each pixel with the same noise. You agree than F stop is important, but disagree with importance of sensor size. Why? Both things have influence on the initial image quality as well as scene illumination.

Further a camera can process the image, increase exposure time, increase AGC, use NR, but initial image quality depends of the sensor size and F-stop. Of course, performance of sensors of different generations can differs. there are better sensors (higher transfer efficiency, lower noise) which produce better initial image with the same pixel size. Therefore it is correct to compare sizes of sensors of the same technology.

I thought Stanislav's comment was totally on topic and excellent. The test is low light performance and Stanislav is commenting on the basic physics of low light performace, how could it be any more on topic?

I'm thinking that Stanislav hijacked the thread. It's okay, that type of thing happens. But let's be clear, just because you see more gain in the resulting image of one camera doesn't mean the other manufacturers had "gain in reserve" (or any other tuning metric) that could have been applied, but they artifically limited it.

Every DSP (Ambarella, Sony, Vatics, Panasonice, etc.) processes the data coming from the imager differently, and every imager reports the data differently. While how to optimize that data is fairly well understood, the end result will still vary. This is true of any electronic family of devices. No two manufacturer's HDTV's function exactly the same, no two manufacturer's digital still cameras produce the same quality images, and certainly no two IP surveillance cameras perform the same.

In my extensive experience, and this goes all the way back to the analog days a decade ago, pretty much any camera can apply far more gain to the image than is practical to use. Gain is just amplification and it not only amplifies the usable video, but it amplifies the distortion (noise). Little post-process tricks like 2DNR and 3DNR can reduce that noise, at the expense of motion blur.

Let's be clear on another topic, increasing gain and thus noise destroys the compression efficiency of a camera. Noise is viewed by almost every CODEC on Earth as motion; and motion and compression efficiency are mortal enemies, even with H.264. In worst-case-scenarios, in a CBR environment, turning up the gain too much will cause extreme compression to be applied (or frames to be dropped), which then kills any advantage you may have gleaned from turning up the gain. If you have hi-cap VBR, then of course the tradeoff is bandwidth and file size, and I know of NO end-users out there that suddenly want their entire array of cameras to spike bandwidth and storage just because it got dark.

There are no magic bullets. Camera manufacturers optimize their hardware to provide a good as possible images for the vast landcape of applications. They all attack it a bit differently to be sure, but to make a statement that "X manufacturer would be better if they just opened up their gain setting." is generally a sign that the person making that claim doesn't really understand how to tune a surveillance camera.

I respectfully disagree. For starters, most cameras won't even apply gain in bright conditions, unless you increase the gain floor, also known as gain bias. this is almost NEVER a preferable position to force the camera into because of the washed out loss of contrast that inevitbly occurs.

Stanislav,

You claim:

"It is impossible to understand why other cameras have not an option to increase AGC gain. It costs zero, only one switch, all cameras must have such option, but with possibility to disable it"

First of all, all of these cameras tested have an 'option to increase AGC gain'. And in these tests, all of them go to their max possible gain setting for each individual camera automatically and by design.

I am dumfounded you continue to make definitive assertions without even a basic understanding of those cameras.

Secondly, there is absolutely a cost in achieving this. IP cameras vary greatly in processing power available to do digital processing that improves low light performance. Plus, manufacturers vary greatly in the expertise and resources dedicated to implement techniques / algorithms to accomplish this.

Net/net, the top performing low light IP cameras have improved greatly and the variance between the 'best' and the 'worst' here has grown.