# Title: A Survey on Video Summarization Techniques
Abstract
- Video summarization methods attempt to abstract the
- Video summarization methods 는 video를 쉽게 해설하기 위해 main occurrences, scenes, or objects in a clip를 바탕으로 요약하는 것.(removing duplications & extracting key frames)
- Video summarization은 frame을 block별로 분할하여 the mean, variance, skew, kurtosis histogram of every block을 계산하고 corresponding blocks of the next frame을 비교한다.
- Key frame extraction하는 다양한 방식이 있으며, 논문 내의 best method는 based on the color distribution 이었다.
Introduction
- 시간적 한계 때문에, 다양한 컨텐츠를 개인이 full contents를 보기 힘드므로 요약하자.
- 비디오 요약은 짧은 시간내에 많은 정보를 함유한 비디오를 생성한다.
- "Video summarization is a vital process that facilitates well-organized storage, quick browsing, and retrieval of large collection of video data without losing important aspects."
- two different kinds of Video Summarization techniques
- static Video summarization: Static video storyboard summary involves a set of key frames from original video and there is no restriction with time and sequence issue.
- dynamic Video Summarization: it selects the most significant, small, dynamic portions of audio and video in order to generate the video summary.
- 대부분의 method는 frame으로 부터 계산된 visual features로 구성.
비디오 요약의 핵심
"In terms of browsing and navigation, a good video abstract will enable the user to get maximum information about the target video sequence in a specified time limitation or adequate information in the minimum time."
| Techniques | Algorithm | Feature | Aim | Frames | Key Frames |
| VSCAN | DBSCAN | Extract color and texture features |
Generating static video summaries | 3266 | 9 |
| VSSUMM1 | K-means | Extract color attributes and visual | Generate dynamic video summaries | 3266 | 15 |
| STILL AND MOVING | K me K-meloids | Extract color and motion vector | Generate dynamic video | 1916 | 19 |
| DT | Delaunary | By capturing the visual contants | Generate Dynamic video summaries | 1917 | 7 |
| VSUHCM | Shot Boundary and Key Frame |
Majority of key frames are extracted | Generate Dynamic video summaries | 3213 | 10 |
Higher Order Color Moments (VSUHCM) : Shot boundary detection and Key frame extraction based on Higher order color moments image Histogram, skewness and kurtosis
Different techniques used for video summarization such as object base summaries, event based summaries, content based summaries, feature based summaries etc.
The methods VSUHCM, DT, STIMO, VSUMM1, VSUMM2 and OV summaries are compared and the comparison analysis is performed based on the color distribution.
Video Summarization using Delaunay Triangulation[1]: 몇 frame을 캡쳐한 후
pre-processing or several trials를 하여 user-specified parameters or prior knowledge을 얻어 최상을 결과 도출.