Reconfigurable Super-resolution Image Sensor

Image sensors are essential components of many imaging systems, like camaras, smartphones, and medical imaging devices. The Pixel readout circuit is responsible for converting the analog charge signal from the pixel into a digital signal that can be processed and stored. The design of the pixel readout circuit is a critical factor in determining the performance of the image sensor, including noise, dynamic range, and readout speed. As the need for the collection and processing of image data continues to rise, the devices and systems involved need to keep evolving to meet performance and cost aspirations. Most of the raw data collected from sensory nodes is unstructured and redundant as such edge devices like cameras, and smartphones with embedded sensory units need to pre-process, temporarily store, and in some applications transfer these data to local or cloud-based processing units. Our work focuses on near-sensor or in-sensor computational imaging, where we apply RTL-based design techniques of both conventional and ML might to extract the most important information from image pixel data with the expectation to improve on latency, privacy security, and energy cost typical of in cloud-based processing.

Conventional Cloud-based AI analysis of image data

Intelligent Image Sensor

The results of this research could have a significant impact on the development of future image sensors. Improved pixel readout circuits could lead to image sensors with better noise, dynamic range, and readout speed. This could enable new imaging applications in areas such as medical imaging, security, autonomous vehicles, and Machine vision.