Browse Prior Art Database

Camera Operation Extraction using 2D Spatio-Temporal Images

IP.com Disclosure Number: IPCOM000114259D
Original Publication Date: 1994-Dec-01
Included in the Prior Art Database: 2005-Mar-28
Document File: 2 page(s) / 56K

Publishing Venue

IBM

Related People

Maeda, J: AUTHOR

Abstract

A technique for extracting camera operations from video data is disclosed that processes many frames at a time. The key technology is texture analysis of two-dimensional spatio-temporal images (hereafter abbreviated to "2DSTimages").

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 60% of the total text.

Camera Operation Extraction using 2D Spatio-Temporal Images

      A technique for extracting camera operations from video data is
disclosed that processes many frames at a time.  The key technology
is texture analysis of two-dimensional spatio-temporal images
(hereafter abbreviated to "2DSTimages").

      A 2DSTimage is a rectangle of which one side lies on the
spatial axis and the other on the temporal axis.  Each side has a
certain length, representing a certain line segment on the screen or
a certain temporal period.  The movement during the period of the
frame contents that lie on the segment can be estimated by
investigating the directivity of the textures in the 2DSTimage.

      Camera operations can be extracted by placing such segments at
appropriate positions in appropriate directions, as shown in Fig. 1:
(a) panning happens when the contents of parallel segments move in
the same direction; (b) zooming in and out are detected when the
contents of radial segments move away from and toward the center of
the screen, respectively.

      A set of successive frames in which the movements of the
contents are almost identical is called a "sub-scene" in this
article.  The procedure for constructing sub-scenes by using the
above technique is as follows:
  1.  An appropriate number of segments are placed in the screen.
       Their positions and directions depend on the motion to be
       extracted.
  2.  Frame sequences are divided at...