In this paper, we propose a new algorithm for fast object detection in the compressed domain of H.264/AVC and HEVC video coding standards. The proposed algorithm is based on three coding parameters: motions vectors, block types and transform coefficients which are extracted from the partially decoded video bitstream. Each feature is separately processed to remove noise and to empowers its discrimination ability. The obtained feature maps are segmented using a fuzzy clustering algorithm. Finally, a fusion step allows merging the segmented feature maps thanks to a weighted linear combination. The originality of this approach lies in its application to both H.264/AVC and HEVC compressed domains. Experiments are conducted using ten test sequences obtained in order to evaluate the performance of the proposed approach. The proposed method has been compared with a state-of-the-art pixel-domain algorithm in term of F-score and computing time.