Jeff Bier wrote in his monthly column that CES 2015 marked the year when the Internet of Things (or what I’d like to call connected devices) began to see.
Whether it’s a sports camera, smartphone or a professional DSLR, most video content of the future will be encoded using formats such as H.265 or VP9, two state-of-the-art compression standards that have been optimized for low bitrate requirements.
However, there is always a long transition period with the introduction of any new codec. Until all devices have the required hardware support for more recent standards, video processors must also support H.264 or VP8; this allows backwards compatibility with existing products for applications like video conferencing.
Since the power budget and memory bandwidth are limited in mobile devices, the huge computation of video encoding is a critical problem for hardware design. Even with the current state of the art CPUs and GPUs, real-time, high-resolution encoding of video in H.265 would not be possible – a full hardware approach is required.
It didn’t happen unless it’s on video
Today we are introducing three new PowerVR Series5 video encoders based on a flexible architecture that scales efficiently from the ultra-low power requirements of wearable devices (e.g. 720p at 30 fps) to the high-performance specifications demanded by markets like medical imaging or professional broadcast (e.g. 4K at 60 fps, 10-bit support):
- PowerVR E5800: targets high-end, professional-level applications (DSLR cameras, entry-level broadcast, medical imaging etc.); this PowerVR encoder is designed to handle 4K videos shot at 60fps; in addition, it features enhanced color depth (up to 10-bits of precision)
- PowerVR E5505: a high-performance encoder targeted at the high-end mobile and embedded market, including smartphones, home entertainment and security cameras; designed to support 4K resolutions at 30fps, it also features 8-bit color with 4:2:0 color rendition and full Android support
- PowerVR E5300: this area-optimized encoder is ideal for wearables, camera-equipped IoT devices, entry-level mobile devices and automotive applications such as ADAS, backup cameras and infotainment; it is designed to process 1080p at 60fps, offering 8-bit color depth and 4:2:0 rendition together with full Android support
All these encoders support multiple standards in a single solution which leads to significant area savings and simplifies system integration; you no longer have to worry about adding several cores to handle multiple formats on the same chip or maintaining multiple drivers.
PowerVR Series5 encoders offer a scalable, easy to integrate solution that takes care of all your video recording needs.
The new PowerVR Series5 video encoders also deliver ultra-low latency modes, with deep sub frame latency achievable from capture to final encode. We’ve implemented dedicated paths for direct input and output which enable the ISP to send metadata and image analytics to the video processor – this reduces latency, memory bandwidth and power consumption, and improves overall performance.
Finally, the PowerVR Series5 video family benefits from a comprehensive driver package that includes RTOS support and multiple rate control options under Android. This will equip smartphone cameras with better options for video recording leading to a more consumer-friendly experience.
Region of interest encoding for better video calls
PowerVR encoders have been proven to deliver better quality than competing solutions at the same bitrate. This is has always been a key advantage of our video IP and an extremely important feature for applications like video conferencing over slow connections.
The improvements in quality in new standards like H.265 will deliver even sharper-looking videos for consumers; in the example below, we are showing a comparison of H.264 and H.265 encoded at the same bitrate – notice the significant improvements in the area around the cactus:
Another significant advantage of our video IP is the ability to work extremely well with other PowerVR multimedia processors. Here is an example of how OEMs can use our PowerVR vision platform to dramatically improve the quality of a typical conference call.
According to scientific research, humans are only able to focus on one area in a frame, which is defined as the region of interest (ROI). When it comes to video conferencing, the region of interest is centered on the face of the user.
This gives the video encoder the information it needs to utilise more bits for the important facial region, whilst using less bits for the background. In this example, the ROI is detected using a face detection algorithm running on a PowerVR Rogue GPU; then the relevant data is passed on to the PowerVR E5500 encoder which uses it to dynamically allocate more bits to the ROI and decrease the level of detail in all non-ROI areas. The use of this algorithm allows the overall bitrate for satisfactory quality video to be reduced, and therefore provides a lower cost, power and bandwidth solution.
We’ll be showing these demos – and many more – at MWC 2015 in Barcelona (hall 6, booth 6E30). Make sure you also follow us on social media (@ImaginationTech, @PowerVRInsider) for the latest news and announcements from Imagination.