Clustering-based Object Detection for Low-resolution Video Streaming

This paper presents a novel strategy for the detection and tracking of objects in low resolution video sequences. The processing is performed in run-time, considering only few buffered frames. Our approach consists of three main steps: (i) spatial segmentation by means of clustering, (ii) candidate set reduction based on feature extraction (iii) choice of the best candidates. The searching region is chosen adaptively, considering the position and features of the object in previous frames. We evaluate the proposed method on a set of diverse soccer sequences. The results indicate excellent performance of our method in comparison to other approaches, especially in terms of robustness against false error detections.