To provide a novel video reconfiguration technology for automatically generating a reconfigured video which maximizes impression impact of a video to be processed.
A feature amount is extracted every frame from image information or sound information which a video to be processed has and the video to be processed is divided into segments on the basis of the amount. Then, the extracted feature amount is vector quantified, converted into the limited number of codes, and the number of converted codes is counted every segment to generate a histogram, so that the histogram is imparted to each segment. Next, a method, such as pLSA (Probabilistic Latent Semantic Analysis), is applied on the basis of the histogram imparted to every segment so as to calculate a probability value vector indicating the degree of attribute to groups of feature amounts behind, to obtain an estimated value of impression impact given to a viewer on the basis of the vector, and to preferentially select a segment which has a great impression impact. Thus, a reconfigured video is determined.
COPYRIGHT: (C)2011,JPO&INPIT
Takashi Sato
Akira Kojima
JP2009111938A | ||||
JP9044639A |
Next Patent: METHOD FOR PRODUCING MATERIAL FOR CIVIL ENGINEERING AND CONSTRUCTION FROM GARBAGE