Resource-Aware Harris Corner Detection based on Adaptive Pruning ⋆ Johny Paul 1 , Walter Stechele 1 Manfred Kr ¨ ohnert 2 , Tamim Asfour 2 Benjamin Oechslein 3 , Christoph Erhardt 3 , Jens Schedel 3 , Daniel Lohmann 3 , and Wolfgang Schr¨ oder-Preikschat 3 1 Technical University of Munich, Germany 2 Karlsruhe Institute of Technology, Germany 3 Friedrich-Alexander University Erlangen-Nuremberg, Germany {johny.paul, walter.stechele}@tum.de {manfred.kroehnert, asfour}@kit.edu {oechslein, erhardt, schedel, lohmann, wosch}@cs.fau.de Abstract. Corner-detection techniques are being widely used in computer vision – for example in object recognition to find suitable candidate points for feature registration and matching. Most computer-vision applications have to operate on real-time video sequences, hence maintaining a consistent throughput and high accuracy are important constrains that ensure high-quality object recognition. A high throughput can be achieved by exploiting the inherent parallelism within the algorithm on massively parallel architectures like many-core processors. However, accelerating such algorithms on many-core CPUs offers several challenges as the achieved speedup depends on the instantaneous load on the processing elements. In this work, we present a new resource-aware Harris corner-detection algorithm for many-core processors. The novel algorithm can adapt itself to the dynamically varying load on a many-core processor to process the frame within a predefined time interval. The results show a 19% improvement in throughput and an 18% improvement in accuracy. Keywords: Harris corner detection, resource-aware programming, invasive com- puting, adaptive pruning 1 Introduction Corner detection is used within computer-vision algorithms like motion detection, image registration, video tracking, feature descriptors for object recognition etc. to infer the contents of an image. Several corner detectors exist today in the literature and comparative evaluations have shown that the Harris [9] corner detectors achieve some of the best results. Recent evaluations in real-time applications such as video tracking [7], visual SLAM [8] and robotic navigation [19] have demonstrated that the ⋆ This work was supported by the German Research Foundation (DFG) as part of the Transre- gional Collaborative Research Centre “Invasive Computing” (SFB/TR 89)