Temperature Dependence of FPN in Logarithmic CMOS Image Sensors

1. Temperature Dependence of FPN in Logarithmic CMOS Image Sensors Dileepan Joseph¹ and Steve Collins² ¹University of Alberta, Canada ²University of Oxford, UK

2. Outline • Image Sensors • CCD versus CMOS • Linear versus logarithmic • Logarithmic Imagers • Fixed pattern noise (FPN) • Colour rendition • Temperature dependence • Conclusion IMTC 2007

3. Image Sensors • A digital camera consists of many components, of which the image sensor is considered principal • Typical imagers may be charge coupled device (CCD) sensors or complementary metal-oxide-semiconductor (CMOS) sensors IMTC 2007

4. CCD Image Sensors • March photo-generated charge systematically from an array of pixels to an output amplifier • Established technology • High resolution, high sensitivity, low noise • Fabrication process is optimised for imaging IMTC 2007

5. CMOS Image Sensors • Work like memory with photosensitive pixels inside each cell • Signal processing may be incorporated on the same die as pixels • High yield and good video performance • May be fabricated by microchip makers IMTC 2007

6. Linear Pixels • Linear pixels (either CCD or CMOS type) “count” photons over a discrete period of time • They produce a voltage directly proportional to the light intensity • Unfortunately, the response may saturate white or black easily © IMS Chips http://www.ims-chips.de/ IMTC 2007

7. Logarithmic Pixels • Logarithmic pixels (CMOS only) measure the “rate” of photon incidence continuously • They produce a voltage directly proportional to the logarithm of the light intensity • The response is similar to that of human vision © IMS Chips http://www.ims-chips.de/ IMTC 2007

8. The Problem • Logarithmic pixels are great for high dynamic range video but… • FPN is worse compared to typical linear pixels • Colours are worse than for typical linear pixels • Impact of temperature on the image quality is poorly understood IMTC 2007

9. The Solution… • IMTC 2001 • We fixed the fixed pattern noise • IMTC 2002 • We improved the colour rendition • IMTC 2007 • We considered temperature IMTC 2007

10. Fixed Pattern Noise • Two photodetectors in the human eye or in a digital camera are not going to be identical • A varying response to light stimulus causes “fixed pattern noise” • The eye uses motion to factor out the FPN; not practical for cameras IMTC 2007

11. Fixed Pattern Noise • Modelling the FPN of logarithmic pixels, we improved calibration • Responses to uniform stimuli were used to define corrections • Our correction reduced the FPN to the same order as the random temporal noise IMTC 2007

12. Colour Rendition • We have shown how to render accurate colours with logarithmic pixels • A colour mapping was defined using images of a reference chart • Perceptual error of the rendered colours was comparable to that of consumer cameras Rendered Reference IMTC 2007

13. Temperature Dependence • Unlike with humans, digital cameras do not regulate temperature • Temperature affects the response of a pixel to a light stimulus • A “new” FPN appears when the temperature dependence varies from pixel to pixel IMTC 2007

14. Temperature Dependence • The dark response of a pixel depends only on temperature • Thus, it may be used to correct FPN due to temperature in the light response • We validated this idea by simulation with real mismatch data IMTC 2007

15. Conclusion • Logarithmic CMOS image sensors are ideal for capturing high dynamic range video • Our research aims to improve the image quality of these cameras from machine grade to consumer grade and better • The dark response of the image sensor may be used to correct temperature-dependent fixed pattern noise in the light response • Future work will simplify our methods and implement them in a complete prototype IMTC 2007

16. Acknowledgements • The authors gratefully acknowledge the support of the Natural Sciences and Engineering Research Council of Canada and the Engineering and Physical Sciences Research Council of the United Kingdom IMTC 2007