Forum on Noise in Imagers on ISSCC 2007

The same prematurily published ISSCC 2007 advance program has interesting forum on imagers noise problems:

IMAGER DESIGN FORUM
F4: Noise in Imaging Systems

Organizer: Albert Theuwissen, DALSA, Eindhoven, Netherlands
Committee: Dan McGrath, Eastman Kodak, Rochester, NY
Jed Hurwitz, Gigle Semiconductor,
Edinburgh, United Kingdom
Hirofumi Sumi, Sony, Tokyo, Japan
Boyd Fowler, Fairchild Imaging, Milpitas, CA
Makoto Ikeda, University of Tokyo, Tokyo, Japan
Takao Kuroda, Matsushita, Kyoto, Japan
Johannes Solhusvik, Micron Technology, Pasadena, CA
Yonghee Lee, Samsung, Gyeonggi-Do, Korea

Noise in Imaging Systems has much in common with noise in the classical world of analog electronics, but imaging adds some very specific noise issues to consider. In many cases the electronic engineer only refers to temporal noise when discussing noise, but in an imaging system non-temporal noise sources also need to be taken into account. In addition, the perfect image sensor in a perfect camera still suffers from noise, because of the photon shot noise of the input signal.

This forum is organized to contribute to a better understanding of noise issues in imaging systems and to stimulate creativity in this field. The speakers at this forum are world experts in this area.

Takao Kuroda (Matsushita) will introduce the topic to begin the forum. The next talk, by Boyd Fowler (Fairchild Imaging), will discuss several noise mechanisms starting with the most important one, kT/C noise. The kT/C noise sets a fundamental detection limit on capacitive sensors. Therefore it is important to understand the factors that determine the kT/C noise and how this noise may be mitigated.

Bedrabata Pain (JPL) will speak on the topic of device-level noise. The pixels in imagers are becoming extremely small and several noise sources can be distinguished within every pixel. Shrinking the CMOS technology will put constraints on the pixel’s noise behavior.

A CMOS imager is usually a complex mixed analog-digital system-on-chip and circuit noise often dominates the total noise of the image sensor. The circuit noise is observed as a fixed pattern noise or a temporal random noise. The former is originated by device mismatches and is cancelled in the analog and/or digital domains, while the latter is more problematic. Shoji Kawahito (Shizuoka U) will discuss noise at the circuit level.

One level higher than the circuit is the system level. Each of the system elements plays an important role in determining the overall noise of the system. The optics may introduce noise in the form of stray signals, such as flare and ghost images, both of which result from internal reflections. A system-level view must also consider factors such as the pixel spectral response, which affects noise amplification, as well as the noise originating from power supply variation, timing jitter, and imperfect FPN cancellation circuits. The speaker invited to talk about system-level noise is Rick Baer (Micron Technology).

Random noise and distortion added to an image signal only matters when it can be seen. Seeing it however does not necessarily mean that it will be considered as a defect. Visual artists learn to use the features of a technology that impart a distinctive look to the resulting image for artistic goals. Jim Larimer (ImageMetrics) will review the properties of the human visual system that allow viewers to see distortion and noise in the temporal, spatial and intensity domains of the image, how the eye samples the signal and how this process can “beat” with capture and reconstruction methods.

The last presentation of the forum will highlight some algorithms used to cancel noise in images. Aleksandra Pizurica (Ghent U) will review some of the latest and best available multiresolution methods for noise reduction. Attention will be given to the following topics: estimation of the noise statistics from the input image (or video); construction of spatially adaptive denoising methods; motion estimation/compensation and noise suppression adapted to motion estimation reliability. In addition, some application specific topics, such as the use of “noise patterns” and camera reference frames for denoising digital camera images will be reviewed.

Although this forum focuses on imaging systems, the issues and techniques dealt with are also applicable to other emerging fields.