CMOS Camera Exposure Adjustment: Principles, Imaging Quality, and Influencing Factors

Principles of CMOS Image Sensor

The target object reflects sunlight onto the sensor, which converts the light signal into an analog voltage signal. Then the analog voltage signal is converted into a digital signal via an ADC, and therefore the digital signal can be processed and used.

image.png

Concepts Related to Image Quality

  • Grayscale: The light intensity information for each pixel, which can be divided into several levels from black to white.
  • DR (Dynamic Range): Each pixel in the image counts the photons received during the reading phase or exposure and outputs a light intensity. Dynamic range describes the ratio between the minimum and maximum measurable light intensity of an image, from pure black to brightest white.

image.png

  • SNR (Signal-to-noise ratio): The incoming light can be seen as a ratio between the grayscale and the noise introduced by signal-processing components. The higher the SNR, the better the image quality, and conversely, the lower the SNR, the lower the image quality. (The high SNR in the figure below is equivalent to the DR.)

image.png

Diagram of the relationship between signal and light intensity

Images with Different Grayscale Values

Image and point cloud with a low SNR:
image.png
Point cloud with a high SNR:
image.png
Oversaturated image and point cloud:
image.png

Imaging Quality and Related Factors

  • Camera Exposure Time: The exposure time refers to the time interval from shutter opening to closing. Usually, the longer the exposure time, the brighter the captured image.
  • Overexposure: When the exposure time is too long, the image will be too bright. The grayscale value of the pixel in the image may reach a maximum of 255, which makes the pixel appear white. In this case, the exposure time should be reduced.
  • Underexposure: When the exposure time is too short, the image will be too dark. The grayscale value of the pixel in the image may reach 0, which makes the pixel appear black. In this case, the exposure time should be increased.
  • Desired Exposure: The grayscale values of pixels range from 180 to 220, and no pixel will reach the maximum grayscale value.
  • Criteria for Point Cloud Quality:
    • Whether there is a point cloud and whether the point cloud is complete
    • Point cloud noise (fluctuation): The point cloud appears irregular. The surface of the point cloud is not flat, and the points in it fluctuate.
  • Summary:
    • The image information is represented by grayscale values (0–255) of each pixel. Both overexposure and underexposure lead to a loss of valid information in the image and will increase the noise, and therefore the SNR will be decreased. For 3D cameras, overexposure and underexposure will cause a loss of valid point clouds or increased noises.

Overexposure Scenarios: For 2D images, overexposure makes the image too bright and appear washed out, and will lead to a loss of point cloud. For 3D images, overexposure causes incomplete point clouds, point fluctuation, and noises.
The figures below are the 2D image, depth map, and point cloud with fluctuation.

Underexposure Scenarios: For 2D images, underexposure makes the image too dark. For 3D images, underexposure causes incomplete point clouds and noises.
The figures below are the 2D image, depth map, point cloud with noises, and normal point cloud.