Corner detection for real-time mobile imaging application

In today's world, mobile devices such as smart phones, tablets and laptops all have integrated cameras to enable video and imaging applications. These devices are typically battery operated, have limited power and hardware resources. Therefore, it is essential to study algorithms that are suitable for implementation under these limitations. In order to reduce cost and power while at the same time providing increased DSP performance, recently there has been a trend to move towards multicore systems in the embedded space. Advanced applications that require sophisticated algorithms have started to appear in mobile devices. In this paper we present how corner detection (interest points) can be parallelized and implemented on the hx3100™ processor, a many core DSP architecture with very low power operation. We also present the algorithmic modifications that enable the real-time, deterministic execution of the code.