Correlation-filter-based trackers still suffer from the problem of spatial restriction and have difficulty in handling unreliable tracking results. In this paper, we propose a robust correlation tracking method by integrating coarse-to-fine redetection scheme and spatial–temporal reliability evaluation strategy. Specifically, we first exploit an improved multi-region coarse estimation method to identify multiple possible candidate regions. Then, the dense search is performed in candidate regions and some candidate objects are obtained. Considering that a good sampling center can help obtain more reasonable response values for the target and other objects, the proposed redetection scheme jointly considers the sampling center quality and detection quality to determine the redetection result. To reduce the risk of drifting to similar distractors in the expanded search area, a novel reliability evaluation strategy is developed with spatial–temporal context to measure the sampling center quality and further elaborate the reliability scores of candidate objects. Furthermore, an effective output integration method is proposed to determine the final output by carefully checking the tracking result and redetection result with historical information and distance constraint. Extensive experimental results demonstrate that the proposed method can achieve favorable tracking performance compared with some state-of-the-art trackers.