Creating a highly efficient motion compensated standards converter without compromising on quality requires careful decisions to be made in processing choices. Find out how InSync Technology created a high quality software frame rate and format converter suitable for deployment on even the most processor-constrained platforms.
Abstract
Broadcasters and content providers know that only motion compensated standards conversion provides top quality results. Historically, high quality motion compensated processing relied on computationally intensive algorithms in order to calculate accurate motion vectors, which implied expensive hardware.
In this whitepaper we show how more efficient motion processing algorithms enable the development of software converters, without compromising on quality. Through a combination of innovative processing methods and careful attention to implementation efficiency, we can create a fully motion compensated standards converter in flexible software, suitable for deployment on any CPU platform, from laptop PC to cloud-enabled data centre.
Efficient processing is needed for many video applications, simply due to the sheer amount of data which needs to be manipulated. With UHD resolutions and higher frame rates, any signal processing device will have to handle in excess of 12Gb of data every second. Applications requiring computationally intensive digital signal processing, such as motion compensated standards conversion, pose a significant challenge because of the enormous quantity of mathematical operations required to obtain precise motion estimation which meets industry-standard picture quality expectations. Minimising these processing requirements to achieve practical implementations becomes even more daunting in software, especially since the preferred solution is platform-agnostic and non- GPU dependent to enable the most flexible deployment for users.
Fortunately, good engineers relish the requirement to create efficient and elegant solutions. Focusing on the constraints of least number of operations, small memory footprint, and multi-platform implementation, whilst maintaining good visual quality leads to innovative solutions for motion vector estimation.
InSync’s proprietary algorithms are carefully tuned to identify redundancy in the video sequences which allow data condensing, enabling later stages of processing to work on smaller sets of data, thereby reducing the amount of computation needed to achieve the desired results. With the right pre-processing, we may confidently reduce the amount of picture information which is passed to the motion estimation stage, thereby saving on a large number of irrelevant calculations.
Lower computational requirements lead immediately to more efficient deployment as software. Taking advantage of best-practice in software implementation, including opportunities to run processes in parallel, enables a motion compensated converter to be deployed as a plug-in suitable for even highly resource constrained platforms such as laptops.
InSync’s FrameFormer software standards converter has been developed as a CPU-only implementation, avoiding the need for GPUs or specialist hardware. A CPU-only implementation also enables easy deployment in cloud configurations. The FrameFormer software standards converter can be run on any platform from low-resource laptop to generic cloud instance, without the need for specialist hardware.
Dr Paola Hobson, Managing Director, InSync Technology
No comments yet