報(bào)告題目：Fast Filtering-based temporal saliency detection using minimum barrier distance
主講嘉賓：王瓊博士 (法國國立應(yīng)用科學(xué)院）
邀請(qǐng)人：鄒文斌老師
時(shí)間：2016年7月20日（周四）上午10:00 -- 11:00
地點(diǎn)：深圳大學(xué)南區(qū)基礎(chǔ)實(shí)驗(yàn)樓北座信息工程學(xué)院N723會(huì)議室
報(bào)告摘要:
In the video salient object detection domain, the optical flowtechnique is widely used to extract temporal saliency. The bigissue with this approach is that the global motion, when it exists,may be wrongly detected as a salient object. To copewith this problem, we propose to filter out the global motionby using the boundary connectivity cue. Firstly, an edgeawarefilter called the guided filter is introduced to preprocessthe color optical flow map for enhancing object edges. Thenthe pixel’s connectivity to the boundary is achieved by usingthe minimum barrier distance in the filtered optical flow map,which leads to the effective removal of the global motion.The proposed method is assessed on the popular Seg-Trackv2 and Fukuchidatasets and then compared to state-of-the-artmethods. The experimental results show that the proposedapproach outperforms the existing related methods.
歡迎各位老師和同學(xué)參加。
