How Exposure Impacts Low Light Video SurveillanceBy IPVM Team, Published on Dec 07, 2009
Exposure, and the tricks manufacturers play with exposure, are key to accurately and fairly assessing low light performance of cameras.
For instance, take these two images:
The images capture the same scene but look substantially different. However, which one is better?
No doubt, the image of the right (Camera B) is much brighter and the vehicle in the background is much easier to see. However, these two images are from the same camera - simply with a change to the maximum exposure setting.
Keep in mind that:
- Using still images as 'proof' of video surveillance quality can easily be used to mislead
- The 'hidden' problem is that long exposure times blur moving objects to the point of being unusable.
To be clear, the answer in our poll above has to be Option C. Not enough information can be determined from a still image. It is not a fair assessment to say that Camera B is a superior product because of it's image quality. In fact, both images are from the exact same camera. The key differentiator between the two shots is that they were captured at different exposure settings - the 'dark' Camera A at 1/30s and the 'bright' Camera B at 1 second.
Exposure Sample Videos
Download the sample video clips that demonstrate a moving subject [56 MBs] with the exposure configurations at 1/60 s, 1/30 s, 1/15 s, 1/8 s, 1/2 s, and 1 s. This should give you a starting point to determine for yourself what trade off between discernible subjects and motion blur is acceptable for your security application.
View the screencast to see the actual video clips compared under different exposure configurations. Note that the subject on the 1 second exposure setting is distorted by a significant amount of motion blur. Is the subject running or walking? If the subject ran fast enough, he would look like a ghost across the screen: that is not an acceptable configuration for monitoring people.
Camera Exposure Settings
In this screencast, we examine the issues and process of adjusting maximum exposure time settings in 5 cameras.
- The defaults for maximum exposure time varies by camera. The Axis Q175, Stardot and Sanyo have 1/30s default for maximum exposure. The Basler has 1s default for maximum exposure (though they report planning to change this to 1/8s in a future release). The Pelco has 1/30s max during day mode operation and 1/8s during night mode. Going from 1/30s to 1s is a 30x difference in the amount of light captured - a significant variance that will have dramatic performance impact.
- These settings generally can only be changed from the camera's web interface as VMS generally do not support adjusting these settings in their administration tools.
- The naming, location and options for each camera's maximum exposure time settings vary. For instance, Axis and Basler provide options in a range of fractional seconds while Sanyo's options are a multiple relative to the capture rate. See the video below for more details.
Risk of "Sleight of Hand" Marketing
Referring back to the poll question example images, it is a trend for some manufacturers to market their camera's night time quality in this fashion. It is important that you factor in the specific settings used to capture the video.
Providing users options for maximum exposure time is a good thing. Marketing great image quality through the use of very long exposure times is a dangerous practice, specifically because of manufacturer's omitting to disclose the problems with motion blur.
Here are our recommendations for maximum exposure time settings:
- For applications requiring a strict 30fps, then the maximum exposure should be no more than 1/30s. Otherwise, the longer exposure makes it impossible to generate a unique image for each frame. That being said, for night time, low light viewing, 30fps is likely to be unnecessary.
- For moving people (running or walking), 1/8s maximum exposure setting should be fine to eliminate motion blur and provide usable images throughout the entire video. While many manufacturers set their maximum at 1/30s, we believe that 1/8s will significantly increase visible details (because of more light) while not introducing new defects.
- For moving vehicles, more than 1/8s and perhaps up to 1/30s maximum exposure settings may be needed (this, of course, depends on the speed of the vehicles).
- If you are willing to sacrifice images of moving objects with the understanding that when they stop you should capture a clear image, than 1/2s or 1s maximum exposure time makes sense. However, this tradeoff needs to be clearly appreciated by the users prior to commissioning the system.