Our new H.264 codec: low delay, 10/12-bit, no DRAM
At the end of September, we announced our new H.264 High Intra encoder and decoder, optimized for automotive and wireless displays. The codec implements the H.264 High 4:4:4 Intra Profile, has very low encoding and decoding delay of <1ms, and supports 8, 10 or 12-bit samples for higher-dynamic-range video. The videantis H.264 High 4:4:4 Intra codec enables compression with virtually no loss of quality, which is of particular importance for safety-relevant automotive vision algorithms, since these will not work well if there are any coding artefacts present. The codec does not require external DRAM memory, reducing cost further when integrated into camera or remote display SOCs. “Videantis has been an excellent partner that addresses our need for high-quality, low-delay video codecs,” said Eric Pinton, Renesas, Senior Manager, Automotive Information Systems.
Show report: Ethernet Technology Day, Stuttgart
The Ethernet Technology Day focuses on bringing low-cost Ethernet AVB wiring to the car, replacing expensive coax or LVDS cabling for cameras. Videantis exhibited at the Renesas booth, showing our low-latency, high-quality H.264 codec integrated with Renesas’ Ethernet AVB solution. At the show, BMW was showing their new Ethernet AVB-based X5, which has been announced, but isn’t commercially available yet. The car includes 5 cameras, for front/back and top-view applications. The latest Mercedes-Benz S-class has 7 cameras on board. These are both high-end models, but cameras are quickly being integrated across the price spectrum, for improved safety and a more comfortable ride. For example, in the US, rear-view cameras will be mandated for new vehicles sold from 2015 onward. While most cameras aren’t H.264-enabled today, there’s a drive toward adoption of H.264 due to the higher quality it provides. We’re looking forward to seeing you all again at next year’s show, which will be held at the heart of the US-based automotive industry: Detroit.
Show report: Embedded Vision Summit
This year’s fall summit was held on October 2nd in the Boston, Massachusetts area. Even though the event wasn’t co-located with the DESIGN East event, like last year’s, attendance grew. Our own Marco Jacobs gave a talk on feature detection and tracking. Videos for all the presentations will be posted online in a couple of weeks on the embedded vision alliance website.
In our booth, we were showing our low power vision accelerator IP core, running diverse vision applications such as face or skin detection, as well as key vision algorithms like feature detection and tracking, all at HD resolution, consuming very little power. Videantis thanks everyone for visiting our booth at the summit and attending our talk. It was great meeting new and old friends in the industry and it’s clear that the computer vision community is thriving.
Interesting industry news
Mercedes-Benz uses stereo cameras to smoothen the ride
“You almost glide over the road, as if you are on a flying carpet,” Mercedes-Benz says. That sounds like magic, but instead Mercedes-Benz uses real-world computer vision technology to accomplish this: the new S-Class car has stereo video cameras to scan and measure road surface undulations ahead. The car’s suspension is then in real-time actively adjusted to smoothly go over the bumps. Another interesting example of how computer vision results in a more comfortable and safer ride.
HEVC doubles compression efficiency, but triples royalties?
For the same picture quality, HEVC reduces the data rate needed by 50 percent when compared to current state-of-the-art H.264. Granted, for real-time, high-volume applications, it will be hard to reach the full 50% reduction, but 30% seems quite doable. That extra bandwidth is sorely needed as more and more people stream higher resolution content to their devices. MPEG LA licenses patent pools that cover the essential compression patents. For HEVC no terms have been announced yet however, and word on the street has it that HEVC royalties will be three times higher than for H.264. Will this hamper HEVC adoption?
Google acquires gesture interface startup Flutter
In contrast to Leap Motion’s external sensor, or Samsung Galaxy S4’s infrared sensors, Flutter uses a Mac’s standard webcam to capture and recognize hand gestures. “We share Google’s passion for 10x thinking, and we’re excited to add their rocket fuel to our journey,” their acquisition statement says. That’s marketing speak for “we’re looking forward to integrate our technology into the Google ecosystem”. Videantis processors run such gesture recognition algorithms in a power-efficient manner, often three orders of magnitude lower than using the host CPU, enabling these technologies to be adopted by high-volume, battery-operated devices.
|ARM TechCon||October 29-31, Santa Clara, USA
|Exhibiting and showcasing our computer vision and video coding technologies|
|Bits & Chips 2013 Embedded Systems||November 7, Netherlands||Presentation by Marco Jacobs: “A vision processing system for a safer, more comfortable, and less expensive ride.”|
|CES||January 7-10, Las Vegas, USA||Join us at the world’s biggest consumer electronics show|
|Mobile World Congress||February 24-27, Barcelona, Spain||Let’s meet at the world’s largest mobile technology congress|
Schedule a meeting with us at these events by sending an e-mail to firstname.lastname@example.org. We’re always interested in hearing about your video and vision ideas and challenges. We look forward to talking with you!