Abstract | ||
---|---|---|
With the rapid development of digital technology, videos are playing an increasingly important role in social networks. Automatically detecting scenes from a video, viewed as an image sequence, is an important task in video analysis. This is usually done via image sequence clustering, the first task of which is representation. Image sequences are very high dimensional and noisy in their form, even though they are intrinsically low dimensional in their content. It is however a challenge to represent image sequences in its intrinsically low dimensional space in a noise-robust way. In this paper, a novel sparse subspace clustering method is proposed for clustering image sequences, which extends the state of the art sparse subspace clustering method by simultaneously performing wavelet multi-scale transform on images and block-diagonal prior spectral clustering based denoising on image sequences. It applies the wavelet multi-scale transform on images based on the spatial similarity of images to reveal the within-image structure, it adds a penalty term in the formulation of the problem to enhance the correlation between neighboring images in the time domain, and it uses denoising to maximize the within-class correlation and minimize the between-class correlation. Experiments using various public datasets, including the ordered face dataset and video scene segmentation dataset, have shown that the proposed method has lower error rate, and is more resilient to noise than other representative sparse subspace clustering methods. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1109/iThings-GreenCom-CPSCom-SmartData.2017.23 | 2017 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData) |
Keywords | Field | DocType |
Sparse Subspace clustering,Wavelet multi-scale transform,Block-diagonal,Image Sequence Clustering | Time domain,Noise reduction,Spectral clustering,Pattern recognition,Computer science,Word error rate,Correlation,Artificial intelligence,Cluster analysis,Wavelet transform,Wavelet | Conference |
ISBN | Citations | PageRank |
978-1-5386-3067-9 | 0 | 0.34 |
References | Authors | |
14 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Liping Chen | 1 | 1 | 0.68 |
Gongde Guo | 2 | 24 | 6.61 |
hui wang | 3 | 76 | 17.01 |