Processing next to the pixel: Lessons from some developments at CEA-Leti

Abstract : Processing the information next to the pixel offers promising possibilities to overcome the classical bottlenecks of image acquisition and processing. CEA-Leti has developed several image sensors with various degrees of parallelism and for various purposes. Indeed, massively parallel processing has been explored either for information extraction or for enhancing the dynamic range of images of "classical" imagers. One well known limitation of such massively processing approach is the limited area available in each pixel. Therefore CEA-Leti has explored the potentials of 3D IC stacking technologies. Yet, in many cases, a high degree of parallelism is not required and processing can be done at the column level. Even at this lower level of parallelism, significant gains in terms of power consumption and/or processing time can be achieved, on occasion by the implementation of emerging signal processing concepts.
Document type :
Conference papers
Complete list of metadatas

https://hal-cea.archives-ouvertes.fr/cea-01837011
Contributor : Léna Le Roy <>
Submitted on : Thursday, July 12, 2018 - 3:52:13 PM
Last modification on : Monday, February 25, 2019 - 4:34:21 PM

Identifiers

Collections

CEA | DRT | LETI | LIST | CEA-GRE

Citation

A. Dupret, A. Verdant, L. Millet, L. Alacoque, A. Peizerat, et al.. Processing next to the pixel: Lessons from some developments at CEA-Leti. 2014 14th International Workshop on Cellular Nanoscale Networks and their Applications (CNNA), Jul 2014, Notre Dame, United States. ⟨10.1109/CNNA.2014.6888641⟩. ⟨cea-01837011⟩

Share

Metrics

Record views

45