This paper proposed a robust and hierarchical camera motion classification approach based on integrated histogram. First, a motion direction histogram is built, and the information entropy of histogram is employed to classify camera motion operations into scaling operation and non-scaling operation. And then, static and translation operations are distinguished by distribu- tion information of histogram, meanwhile the direction of translation operation is also identified. Experimental results show that the new approach can efficiently achieve global motion estimation.
A fusion architecture based on TBM for camera motion classification
