LONDON DSP core licensor Ceva Inc. is due to unveil a software-programmable multimedia video processor architecture at the Mobile World Congress in Barcelona next week.
The multicore architecture, called MM3000, which comes complete with C compilers, power management provision and an RTOS/multithreading scheduler is intended to be able to process any and all video codecs up to the highest resolutions and frame rates currently available as well as future codecs for things like 3-D video.
In addition the architecture has been constructed with video analysis in mind, including such things as: augmented reality, gesture recognition, face detection and recognition, facial expressions recognition and object detection.
This is a major undertaking and the result of four years work, said Eran Briman, vice president of marketing at Ceva. "For most cell phones the video processing is a requirement but video analytics in real-time is a way OEMs can differentiate," Briman said. In the area of video codecs there are many standards, with more on the horizon, Briman said. In the area of video post-processing there are no standards, he added. Either way it makes software programmability a necessity but so too is power efficiency.
And lest it thought that recognizing emotions in a video captured with a smart phone is an unnecessary sophistication Ceva conceives of the MM3000 as a video processing engine for every high-definition application, from smart phones, via computers to games consoles and digital televisions.
At MWC Ceva plans to demonstrate the MM3000 architecture implemented in an Altera FPGA with a Nios processor providing housekeeping support.