• The Journal of Korean Institute of Information Technology
• /
• v.17 no.5
• /
• pp.13-20
• /
• 2019

Recently, object detection technology based on CNN has been actively studied. Object detection technology is used as an important technology in autonomous vehicles, intelligent image analysis, and so on. In this paper, we propose a rotation change robust object detection system by applying MI-FL (Moment Invariant-Feature Layer) to SSD (Single Shot Multibox Detector) which is one of CNN-based object detectors. First, the features of the input image are extracted based on the VGG network. Then, a total of six feature layers are applied to generate bounding boxes by predicting the location and type of object. We then use the NMS algorithm to get the bounding box that is the most likely object. Once an object bounding box has been determined, the invariant moment feature of the corresponding region is extracted using MI-FL, and stored and learned in advance. In the detection process, it is possible to detect the rotated image more robust than the conventional method by using the previously stored moment invariant feature information. The performance improvement of about 4 ~ 5% was confirmed by comparing SSD with existing SSD and MI-FL.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.6
• /
• pp.1703-1712
• /
• 1999

In the paper, we propose efficient article and scene change detection algorithms to make the index of sports news compressed in MPEG-2 domain. In the proposed algorithm, the information in MPEG-2 compressed domain is directly used without decoding to save the computation time. The scene change detection algorithm is constructed in an hierarchical method so that the time for detection can be greatly reduced. Also, the algorithm can provide the robust detection against abrupt illuminance change because the luminance and chrominance components are simultaneously considered. Also, the scene change caused by special effect such as dissolve and wipe can be detected in the compressed domain. In the article detection, the algorithm is constructed for robust detection of the anchor frame using the concept of CCV(Color Coherent Vector).

• International Journal of Contents
• /
• v.4 no.2
• /
• pp.24-28
• /
• 2008

In this paper, we propose a new approach for content-based video retrieval using non-parametric based motion classification in the shot-based video indexing structure. Our system proposed in this paper has supported the real-time video retrieval using spatio-temporal feature comparison by measuring the similarity between visual features and between motion features, respectively, after extracting representative frame and non-parametric motion information from shot-based video clips segmented by scene change detection method. The extraction of non-parametric based motion features, after the normalized motion vectors are created from an MPEG-compressed stream, is effectively fulfilled by discretizing each normalized motion vector into various angle bins, and by considering the mean, variance, and direction of motion vectors in these bins. To obtain visual feature in representative frame, we use the edge-based spatial descriptor. Experimental results show that our approach is superior to conventional methods with regard to the performance for video indexing and retrieval.

• The KIPS Transactions:PartB
• /
• v.9B no.3
• /
• pp.343-350
• /
• 2002

A scene change detection which involves finding a cut between two consecutive shots is an important step for video indexing and retrieval. This paper proposes an algorithm for fast and accurate detection of abrupt scene changes in an MPEG compressed domain with minimal decoding requirements arid computational effort. The proposed method compares two successive DC images of I-frames for finding the GOP (group of picture) which contain a scene change and uses macroblock-coded type information contained in B-frames to detect the exact frame where the scene change occurred. The experiment results demonstrate that the proposed algorithm has better detection performance, such as precision and recall rate, than the existing method using all DC images. The algorithm has the advantage of speed, simplicity and accuracy. In addition, it requires less amount of storage.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1998.06a
• /
• pp.143-146
• /
• 1998

본 논문에서는 MPEG-2 압축영상에서 뉴스 색인을 위한 개선된 장면(Shot)전환 알고리즘을 제안한다. 제안된 검출 알고리즘은 압축된 비디오 시퀀스에서 추출한 특징정보인 에지정보와 컬러정보를 함께 사용하여 GOP단위의 빠르고 효율적인 장면전환 검출 결과를 얻을 수 있다. 제안된 알고리즘은 장면전환 검출뿐만 아니라 내용 기반(content - based) 비디오 검색 등에도 효과적으로 이용될 수 있다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10b
• /
• pp.268-270
• /
• 2001

히스토그램을 이용한 장면 전환 검출 기법은 순차적으로 접근하여 모든 프레임의 히스토그램을 구하고 각 히스토그램의 차를 이용하여 장면 전환을 검출한다. 하지만, 장면 전환이 비교적 적게 나타나는 부분에서는 모든 프레임을 비교한다는 것은 비효율적이다. 본 논문에서는 모든 프레임을 순차적으로 비교하지 않고 가중치를 조절하여 장면 전환이 거의 발생하지 않는 경우는 많은 프레임을, 장면 전환이 많은 곳에서는 적은 프레임을 생략하여 히스토그램을 비교하는 방법을 제안한다. 이 방법은 생략하는 프레임 수를 조절하기 때문에 순차적으로 처리하는 것보다 빠른 처리 시간을 보일 수 있다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.8B
• /
• pp.1438-1443
• /
• 1999

In this paper, we propose detection methods for gradual scene changes such as dissolve, pan, and zoom. The proposal method to detect a dissolve region uses scene features based on spatial statistics of the image. The spatial statistics to define shot boundaries are derived from squared means within each local area. We also propose a method of the camera motion detection using four representative motion vectors in the background. Representative motion vectors are derived from macroblock motion vectors which are directly extracted from MPEG streams. To reduce the implementation time, we use DC sequences rather than fully decoded MPEG video. In addition, to detect the gradual scene change region precisely, we use all types of the MPEG frames(I, P, B frame). Simulation results show that the proposed detection methods perform better than existing methods.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.10
• /
• pp.85-92
• /
• 2023

In this paper, we propose a method for automatically generating story-based abbreviated summaries from long-form dramas and movies. From the shooting stage, the basic premise was to compose a frame with illusion of depth considering the golden division as well as focus on the object of interest to focus the viewer's attention in terms of content delivery. To consider how to extract the appropriate frames for this purpose, we utilized elemental techniques that have been utilized in previous work on scene and shot detection, as well as work on identifying focus-related blur. After converting the videos shared on YouTube to frame-by-frame, we divided them into a entire frame and three partial regions for feature extraction, and calculated the results of applying Laplacian operator and FFT to each region to choose the FFT with relative consistency and robustness. By comparing the calculated values for the entire frame with the calculated values for the three regions, the target frames were selected based on the condition that relatively sharp regions could be identified. Based on the selected results, the final frames were extracted by combining the results of an offline change point detection method to ensure the continuity of the frames within the shot, and an edit decision list was constructed to produce an abbreviated summary of 62.77% of the footage with F1-Score of 75.9%

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.1
• /
• pp.141-148
• /
• 2011

Most of the previous works on scene change detection algorithm focus on the detection of abrupt rather than gradual changes. In general, gradual scene change detection algorithms require heavy computation. Some of those approaches don't consider the error factors such as flashlights, camera or object movements, and special effects. Many scenes change detection algorithms based on the histogram show better performances than other approaches, but they have computation load problem. In this paper, we proposed a scene change detection algorithm with fast and accurate performance using the vertical and horizontal blocked slice images and their macro block informations. We apply graph cut partitioning algorithm for clustering and partitioning of video sequence using generated spatio-temporal histogram. When making spatio-temporal histogram, we only use the central block on vertical and horizontal direction for performance improvement. To detect camera and object movement as well as various special effects accurately, we utilize the motion vector and type information of the macro block.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2011.01a
• /
• pp.51-53
• /
• 2011

최근에는 텍스트정보 보다 동영상정보를 다루는 일이 많아졌고 그에 따라 동영상 데이터의 분할, 색인, 검색 등을 위해 장면 전환 검출이 필요하게 되었다. 장면 전환 검출 기술은 비디오 데이터의 장면 변화가 발생하는 경계를 검출하는 기술이다. 본 논문에서는 에지 이미지를 이용한 장면전환 검출과 이를 위한 임계값 설정, 그리고 결과에서 중복된 이미지와 오 검출 된 이미지를 줄여줄 수 있는 구현에 대하여 실험결과를 바탕으로 설명한다.