Stacked Mosaic Detection based on Line Expansion

When uploading multimedia data such as photos or videos on network, they might be covered with mosaic blocks since some private information is not suitable for exposing. In some situation, mosaic images or video frames need to be restored or trimmed from them, it is necessary to detect the mosaic so as to discriminate these data for subsequent processing. Up to now, most mosaic detection methods are based on edge detection. However, mosaic regions with similar colors may cause several similar mosaic blocks to be stacked together, which makes it difficult to extract sufficient edge features, and lead to hardly detecting the subsequent mosaic region. In this paper, a line expansion- based mosaic detection method is proposed. First, find the upper left point of a block, expand downward to obtain a vertical line. Second, expand to the right based on vertical line to get the whole rectangular block. Third, repeat above steps until the entire mosaic region is extracted, ignoring the regions that have been marked during extraction. The proposed method is compared with several representative methods under different sizes and stacked mosaic. The experimental results show that the proposed method outperforms the state-of-the-art methods on stacked mosaic.