A novel generated method for high quality images

摘要: Thanks to the diversity and popularity of image acquisition technique, data driven methods for analysis edit become popular. However, explosive growth images also presents challenges. To help users retrieve expected quickly effectively is one hardest tasks. It difficult express user's intention completely a single cue, thus combining multiple cues becomes necessary. Another problem that content-based matching process takes long time in case massive data. In this dissertation, we propose progressive retrieval method high quality images, targeting on problems existed cannot guarantee consuming feature material. We start from fastest text strategy used evaluate both clarity independence. Finally, scene elements these object extracted, contours are matched between user sketch candidate elements. parallelize whole framework MapReduce further improve performance. validate performance, accuracy algorithm result various experiments. The results show our can efficiency control better. Then rapidly create through simple line sketch.