Noise is one of the key performance characteristics of any solid-state image sensor. It sets the signal-to-noise ratio and the dynamic range, and it sets the minimum amount of light needed to capture an acceptable image. But noise shows up in many forms. It can be fixed-pattern noise in the spatial domain or temporal noise in the time domain, and it can appear at pixel, column or even row level. That makes noise analysis one of the most complex parts of working with image sensors.
In this 3.5-hour online masterclass, Albert Theuwissen explains where each type of noise comes from and what can be done to keep it as low as possible. Some measures lie in the design and manufacturing of the sensor itself, while others are in the hands of camera engineers, through cooling and camera electronics design. As Albert puts it, hardly anyone has ever complained that the noise in an image sensor is too low.
The masterclass also looks at the positive side of noise. Photon shot noise, for example, gives engineers a simple but powerful tool for analysing and debugging sensor performance, and it can even help to increase the speed of analog-to-digital converters. Finally, he explores how the perception of noise depends on who is looking at the image, a human or a computer, and how the characteristics of the human visual system can inform the camera's signal processing chain.
Topics to be discussed include the origins of noise in CMOS image sensors, fixed-pattern versus temporal noise, pixel, column and row effects, reducing noise through sensor design, cooling and camera electronics, the constructive use of photon shot noise for analysis, debugging and faster ADCs, and how human and machine perception affect the way noise is judged.
Online masterclass with Albert Theuwissen