Scene duplicate detection based on the pattern of discontinuities in feature point trajectories

Abstract

The paper is aiming to detect and retrieve videos of the same scene (scene duplicates) from broadcast video archives. Scene duplicate is composed of different pieces of footage of the same scene, the same event, at the same time, but from the different viewpoints. Scene duplicate detection would be particularly useful to identify the same event reported in different programs from different broadcast stations. The approach should be invariant to viewpoint changes. We focused on object motion in videos and devised a video matching approach based on the temporal pattern of discontinuities obtained from feature point trajectories. We developed an acceleration method based on the discontinuity pattern, which is more robust to variations in camerawork and editing than conventional features, to dramatically reduce the computation burden. We compared our approach with an existing video matching method based on the local feature of keyframe. The spatial registration strategy of this method was also used with the proposed approach to cope with visually different unrelated video pairs. The performance and effectiveness of our approach was demonstrated on actual broadcasted videos.

Scene Duplicate Detection from News Videos Using Image-Audio Matching Focusing on Human Faces

“…This is to prevent image feature matching between different facial parts. Next, considering the possible occlusion by microphones and so on, the similarity of each sub-region pair except for the bottom sub-region is calculated using Wu et al's method [5].…”

“…This method, however, cannot be applied to shots without flashes. On the other hand, Wu et al proposed a method based on the discontinuity of feature point trajectories [5]. This method evaluates the speed and the change of motion of a subject, which makes it robust to the difference of viewpoints.…”

Scalable detection of partial near-duplicate videos by visual-temporal consistency

“…Broadly, we can categorize the existing works into three main groups: signature-based [6,4], keyframe-based [20,22,26,27] and trajectory-based [24,16]. The definition of near-duplicate indeed varies depending on the target application.…”

“…For instance, trajectories have been utilized to highlight different motion behaviors [16] and then assign behavioral labels to each local descriptors. In another recent work [24], the whole shot is represented using a bag of trajectories where each trajectory in turn is described as temporal patterns of discontinuities. In general, the extraction of trajectories is an extremely expensive operation.…”

ViCopT: a robust system for content-based video copy detection in large databases

“…A main difference is that the goal is to find objects or special queries in the video and not on finding copies; the temporal behavior of points is not considered during retrieval. In [21], the authors use object motion in videos and the discontinuities of feature point trajectories to match videos. The goal is to detect the same scenes from different points of view; it is more detecting NDD than copies.…”