LONDON – Imagination Technologies Group plc has said it is "well advanced" in developing camera-related IP cores that can address growing markets in vision processing.
The concept of putting an application-specific processor close to an image sensor to perform tasks related to image capture and extracting useful information has been previously pursued by a number of startup and system companies.
One of the benefits is that in distributed and complex embedded systems it can reduced power consumption and the burden on the central processing unit if raw data is converted into relevant information at the camera. Embedded vision in robotics and such things as security cameras and automobiles are all growing markets for smart cameras that can perform such tasks as pattern-recognition, or recognizing pedestrians or faces.
In 2012 Freescale Semiconductor Inc. announced it had licensed image processing intellectual property from CogniVue Corp. (Gatineau, Quebec) for use in processors for automotive driver assistance systems. Cognivue's Apex image processing cores include multiple parallel computational units – as many as 96 – running in parallel with an ARM processor core used for housekeeping.
In an interim management statement Imagination said that PowerVR Vision would exploit the technological synergies that exist with the company's graphics and video IP cores.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.