Adjust camera exposure: The working principle of CMOS, image quality, and quality influencing factors

The working principle of CMOS in cameras

The object reflects sunlight onto the CMOS sensor, which then converts the light signals into analog voltage signals. Then, an analog-to-digital converter (ADC) converts these analog voltage signals into digital signals, which would be used in processing images afterward.

Image quality terminology

  • Grayscale: A grayscale image is one in which the value of each pixel is a single sample representing only an amount of light. The contrast ranges from black at the weakest intensity to white at the strongest.
  • Dynamic range (DR): Every pixel in the image counts the number of photons it receives during access or exposure, and each number represents a value of light intensity. Dynamic Range (DR) is the ratio between the largest and smallest light intensity values.
  • Signal-to-noise ratio (SNR): SNR is the ratio of reflected light, represented by the grayscale value, to the noise power of signal-processing components. A high SNR indicates a high-quality image, while a low SNR suggests an unqualified image. In the picture below, the high SNR parts correspond to the dynamic range.

A diaphragm showing the relationship between the signal and light intensity:

Light intensity principle in camera imaging (different grayscale values)

Image and point cloud screenshot (a low SNR):

Image and point cloud screenshot (a high SNR):

Image and point cloud screenshot (over-saturated):

Imaging quality and its influencing factors

  • Camera exposure time: In photography, exposure time is the length of time that the film or digital sensor inside the camera is exposed to light (that is, when the camera’s shutter is open) when taking a photograph. Longer exposure time means too much light is let into the camera, which results in an overly bright image.
  • Overexposure: Longer exposure time results in overly bright images, in which many pixels appear white, and the grayscale values of these pixels are 255, the maximum. Therefore, exposure should be decreased.
  • Underexposure: Shorter exposure time results in overly dark images, in which many pixels appear black, and the grayscale values of these pixels are 0, the minimum. Therefore, exposure should be increased.
  • A good exposure: The grayscale value of a good exposure should range between 180 and 220, less than the maximum value.
  • Point cloud quality:
    • Whether point clouds are defective?
    • Fluctuations and noise of point cloud: point clouds appear irregular and uneven, with fluctuations.
  • Summary:
    • The grayscale value, ranging from 0 to 255, makes up an image’s information. Overexposure and underexposure both result in information loss and more noise, leading to a lower SNR. For 3D cameras, problems are defective point clouds and more noise.

Overexposure: 2D flash images are overly bright and white, with defective point clouds. In 3D images, fluctuations and noise occur, with defective point clouds.
Below are a 2D (flash) image, a depth map, and a point cloud screenshot (with fluctuations) respectively:

Underexposure: 2D flash images are dark and black, while in 3D images, noise occurs as well, with defective point clouds.
Below are a 2D (flash) image, a depth map, one point cloud screenshot (with noise), and another point cloud screenshot (normal) respectively:

1 Like