Super-Resolution Imaging Applied to Moving Targets in High Dynamic Scenes

摘要

In modern tracking systems the ability to obtain high quality, high resolution appearance of the tracked target is often highly desirable. However, the reality of operational deployment often means that imaging systems deployed for this task suffer from limitations reducing effective image quality. These limitations can be attributed to a range of causes such as low quality video sensors, system noise, high target dynamics and other environmental noise factors. Despite the advantages of the super-resolution techniques the problem of handling complex motion still remains a challenging task for the effective super-resolution implementation. The computational complexity and large memory requirements required for the implementation of super-resolution imaging largely restrict the usage of these techniques in real-time hardware implementations. In order to improve visual quality of the tracked target and overcome these limitations, we propose a simple yet effective solution that integrates a super-resolution imaging approach based on combination of the Sum of the Absolut Differences (SAD) and gradient-descent motion estimation techniques into a novel tracking approach. In addition, the proposed approach demonstrates robustness in improved target appearance modeling that assists the overall tracking system. The presented results demonstrate this significant improvement in visual target representation whilst tracking over high dynamic scenes. The implementation simplicity of the proposed approach makes it an attractive solution for realization on low power hardware. Such a system can be deployed on small unmanned aerial vehicles (UAV) or other hardware where size, weight and power (SWaP) is of a particular concern.