Automatic Generation of Spatial and Temporal Memory Architectures for Embedded Video Processing Systems
- PDF / 668,551 Bytes
- 10 Pages / 600.032 x 792 pts Page_size
- 66 Downloads / 213 Views
Research Article Automatic Generation of Spatial and Temporal Memory Architectures for Embedded Video Processing Systems H˚akan Norell, Najeem Lawal, and Mattias O’Nils Electronics Design Division, Department of Information Technology and Media, Mid Sweden University, 851 70 Sundsvall, Sweden Received 1 May 2006; Revised 11 October 2006; Accepted 15 October 2006 Recommended by Heinrich Garn This paper presents a tool for automatic generation of the memory management implementation for spatial and temporal realtime video processing systems targeting field programmable gate arrays (FPGAs). The generator creates all the necessary memory and control functionality for a functional spatio-temporal video processing system. The required memory architecture is automatically optimized and mapped to the FPGAs’ memory resources thus producing an efficient implementation in terms of used internal resources. The results in this paper show that the tool is able to efficiently and automatically generate all required memory management modules for both spatial and temporal real-time video processing systems. Copyright © 2007 H˚akan Norell et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.
INTRODUCTION
In today’s society it is apparent that video systems are generally becoming standard applications. These systems are playing a central role in the daily life of the majority of homes. Embedded video applications contained in home entertainment systems are becoming more and more complex as is the processing required. Real-time streamed video is common and this places significant constraints on the processing applications as the data rates increase towards high-definition television (HDTV). Surveillance application is one of the most rapidly developing areas. Video monitoring is now present in almost every store or public place. The amount of video data produced by these systems requires them to be able to efficiently derive features or events that are present in the scene. This, in turn, has led to an increased requirement to enable more complex operations such as prefiltering, object recognition, or compression to be performed as close as possible to the video source. This advance has led to the rapid development of smart cameras which have the ability to fulfill these requirements. Complex reconfigurable embedded systems with advanced image processing functions must be rapidly developed which are also available at a low cost [1]. Video processing systems required in the broadcast and postprocessing market are typically in the low-volume and high-cost segment. These are systems performing real-time high-resolution (2048 2048@24 fps) high-performance
computation with complex motion compensating spatiotemporal filters [2]. The trend is for the algorithm complexity to increase over time. This increased complexity is often reflected in the memory usage. One example invol
Data Loading...
