Time filter

Source Type

Sun T.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques | Jiang X.,Shanghai JiaoTong University | Jiang X.,National Engineering Laboratory on Information Content Analysis Techniques | And 2 more authors.
Elektronika ir Elektrotechnika | Year: 2013

A new classification algorithm of human action recognition of video content is suggested in our paper. It analyzes the variation of the content of video scenes or human action from video bi-modal features, in order to cognize content efficiently and precisely. This scheme is based on the pattern analysis of spatio-temporal slices and audio signature feature extracted from the video files. The Spatial-temporal Variation Histogram feature is firstly defined in our paper. It is applied to describe spatio-temporal variation analysis of human action or video scenes. The audio signature is also applied to identify the audio content by extracting unique signatures from a part of audio signal. The experiments show excellent performance of classification on the KTH dataset.


Tang F.,Shanghai JiaoTong University | Lu H.,Shanghai JiaoTong University | Sun T.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques | And 2 more authors.
2012 5th International Congress on Image and Signal Processing, CISP 2012 | Year: 2012

Image representation and classifier are playing key roles in image classification. An effective combination of image representation and classifier could raise the accuracy of image classification. A novel image classification algorithm based on sparse coding and random forest is proposed in this paper. Sparse coding is adopted to train a dictionary of visual words and then to convert SIFT descriptors into sparse vectors. Afterward several pooling methods and spatial partition are used to pool these sparse vectors to represent images. Random forest, an efficient multiclass classifier, is employed to classify the sparse vectors of images. The outcome of the experiments demonstrates that the proposed algorithm outperforms the state-of-the-art in image classification using Caltech-101 and Scene-15 datasets. © 2012 IEEE.


Chao J.,Shanghai JiaoTong University | Jiang X.,Shanghai JiaoTong University | Jiang X.,National Engineering Laboratory on Information Content Analysis Techniques | Sun T.,Shanghai JiaoTong University | And 2 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2013

In this paper, a novel video inter-frame forgery detection scheme based on optical flow consistency is proposed. It is based on the finding that inter-frame forgery will disturb the optical flow consistency. This paper noticed the subtle difference between frame insertion and deletion, and proposed different detection schemes for them. A window based rough detection method and binary searching scheme are proposed to detect frame insertion forgery. Frame-to-frame optical flows and double adaptive thresholds are applied to detect frame deletion forgery. This paper not only detects video forgery, but also identifies the forgery model. Experiments show that our scheme achieves a good performance in identifying frame insertion and deletion model. © 2013 Springer-Verlag.


Cheng D.,Shanghai JiaoTong University | Sun T.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques | Sun T.,New Jersey Institute of Technology | And 2 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2013

In recent researches, image classification of objects and scenes has attracted much attention, but the accuracy of some schemes may drop when dealing with complicated datasets. In this paper, we propose an image classification scheme based on image sparse representation and multiple kernel learning (MKL) for the sake of better classification performance. As the fundamental part of our scheme, sparse coding method is adopted to generate precise representation of images. Besides, feature fusion is utilized and a new MKL method is proposed to fit the multi-feature case. Experiments demonstrate that our scheme remarkably improves the classification accuracy, leading to state-of-art performance on several benchmarks, including some rather complicated datasets such as Caltech-101 and Caltech-256. © 2013 Springer-Verlag.


Xu K.,Shanghai JiaoTong University | Jiang X.,Shanghai JiaoTong University | Jiang X.,National Engineering Laboratory on Information Content Analysis Techniques | Sun T.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques
Proceedings - International Conference on Image Processing, ICIP | Year: 2015

A novel method for human action recognition is proposed in this paper. Traditional spatial-temporal interest point detectors are easily affected by hair, face, shadow, clothes texture or the shake of camera. Inspired by the use of points distribution information, we propose a point selection method to select representative points (denoted by the 'pose points'), which use HOG human detector and contour detector to select the points on human pose edges. The pose points carry both local gradient information and global pose information. 3D-SIFT scale selection method and novel descriptors called body scale and motion intensity feature are also studied. The descriptors calculate the width scale of different levels of human body and count motion intensity of activity in five directions. The descriptors combine spatial location with the moving intensity together and are used for further classification with SVMs. Experiments have been conducted on benchmark datasets and show better performance than previous methods, which achieved 99.1% on Weizmann dataset and 95.8% on KTH dataset. © 2015 IEEE.


He P.,National Engineering Laboratory on Information Content Analysis Techniques | Jiang X.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques | Wang S.,National Engineering Laboratory on Information Content Analysis Techniques
Journal of Visual Communication and Image Representation | Year: 2016

Videos captured by stationary cameras are widely used in video surveillance and video conference. This kind of video often has static or gradually changed background. By analyzing the properties of static-background videos, this work presents a novel approach to detect double MPEG-4 compression based on local motion vector field analysis in static-background videos. For a given suspicious video, the local motion vector field is used to segment background regions in each frame. According to the segmentation of backgrounds and the motion strength of foregrounds, the modified prediction residual sequence is calculated, which retains robust fingerprints of double compression. After post-processing, the detection and GOP estimation results are obtained by applying the temporal periodic analysis method to the final feature sequence. Experimental results have demonstrated better robustness and efficiency of the proposed method in comparison to several state-of-the-art methods. Besides, the proposed method is more robust to various rate control modes. © 2015 Elsevier Inc. All rights reserved.


He P.,Shanghai JiaoTong University | He P.,National Engineering Laboratory on Information Content Analysis Techniques | Sun T.,Shanghai JiaoTong University | Sun T.,National Engineering Laboratory on Information Content Analysis Techniques | And 4 more authors.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

In this paper, we propose a novel scheme to detect double MPEG-4 compression with block artifact analysis. An adaptive measurement of block artifact in decompressed frames is proposed and then combined with the Variation of Prediction Footprint (VPF) in an effective way. Based on such measurement, periodic analysis is used to detect double compression. The proposed scheme is verified on several publically available standard videos and compared with the state-of-the-art method. Experimental results demonstrate that it has more robust detection capability. © Springer International Publishing Switzerland 2015.

Loading National Engineering Laboratory on Information Content Analysis Techniques collaborators
Loading National Engineering Laboratory on Information Content Analysis Techniques collaborators