We are interested in studying the electronic image sensors that are used in so many of the imaging devices around us. In particular, we are focused on studying all aspects of the defects that develop in these sensors. Defects are typically recognized by casual camera users as blemishes in the final photograph and will eventually limit the usefulness of a camera.

Why study pixel defects?

Objectives of this research

We aim to advance the defect tolerance of solid-state image sensors by

1. Performing laboratory experiments to identify the growth of in-field detects and quantifying their behaviour in real devices.
2. Developing algorithms that automatically detect defects and effectively enable healing in cameras without user intervention.
3. Implementing novel pixel architectures in hardware to avoid or mitigate certain defect types before they can affect image quality.

What are in-field defects?

Pixel defects in image sensors have been discussed for some time. Until now, manufacture-time defects, those that develop during the semiconductor fabrication process and other stages of manufacturing, have been most studied and are reasonably well understood. However, in-field defects are becoming an increasing concern as digital cameras are deployed for longer periods of time in many more applications.

In-field defects:

• Develop after the camera enters service in the field.
  
• Are distinct from manufacture-time defects that may be "hidden" by factory calibration.
• Are difficult to eliminate without factory calibration.
• Have been reported anecdotally by many photographers and other users.
• Have not been quantitatively studied in terrestrial environments.

Documented defect types

From our own characterization of many consumer digital cameras, we have identified some commonly-found defect types: 1. Hot pixels (most common) Respond normally to illumination Add an offset that builds up with exposure duration Become bright after long exposure photos (e.g. in low light) Result in reduced dynamic range
2. Partially-stuck pixels Fixed partial offset May response to illumination Like an extreme hot pixel     3. Fully-stuck pixels Permanently stuck at some extreme value Don't response to illumination Most distracting in final image       4. Abnormal sensitivity defects Respond to light at a different rate than good pixels. Typically calibrated out
5. RTS (Random Telephgraph Signal) defects Behaviour usually like that of a hot-pixel Hotness (dark current) alternates between two values