2018/08/14 by Geethu Miriam Jacob, Jacob, Geethu Miriam, Sukhendu Das +1
Computer Science · #Advanced Image and Video Retrieval Techniques #Biometric Identification and Security #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image and Video Stabilization
paper · pdf · doi:10.48550/arxiv.1808.04551
openalex publication_date 2018/08/14 · openalex created_date 2022/08/04 · openalex updated_date 2026/07/28
Moving Object Segmentation is a challenging task for jittery/wobbly videos.\nFor jittery videos, the non-smooth camera motion makes discrimination between\nforeground objects and background layers hard to solve. While most recent works\nfor moving video object segmentation fail in this scenario, our method\ngenerates an accurate segmentation of a single moving object. The proposed\nmethod performs a sparse segmentation, where frame-wise labels are assigned\nonly to trajectory coordinates, followed by the pixel-wise labeling of frames.\nThe sparse segmentation involving stabilization and clustering of trajectories\nin a 3-stage iterative process. At the 1st stage, the trajectories are\nclustered using pairwise Procrustes distance as a cue for creating an affinity\nmatrix. The 2nd stage performs a block-wise Procrustes analysis of the\ntrajectories and estimates Frechet means (in Kendall's shape space) of the\nclusters. The Frechet means represent the average trajectories of the motion\nclusters. An optimization function has been formulated to stabilize the Frechet\nmeans, yielding stabilized trajectories at the 3rd stage. The accuracy of the\nmotion clusters are iteratively refined, producing distinct groups of\nstabilized trajectories. Next, the labels obtained from the sparse segmentation\nare propagated for pixel-wise labeling of the frames, using a GraphCut based\nenergy formulation. Use of Procrustes analysis and energy minimization in\nKendall's shape space for moving object segmentation in jittery videos, is the\nnovelty of this work. Second contribution comes from experiments performed on a\ndataset formed of 20 real-world natural jittery videos, with manually annotated\nground truth. Experiments are done with controlled levels of artificial jitter\non videos of SegTrack2 dataset. Qualitative and quantitative results indicate\nthe superiority of the proposed method.\n