Browse Prior Art Database

Method of Detecting Scene Changes in Moving Pictures

IP.com Disclosure Number: IPCOM000107604D
Original Publication Date: 1992-Mar-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 3 page(s) / 84K

Publishing Venue

IBM

Related People

Ioka, M: AUTHOR

Abstract

This article describes a method of detecting scene changes in moving pictures. A scene change means a change of camera position from one frame to the next. The inclusion of new objects or the exclusion of old objects in a frame does not constitute a scene change. Camera movements, such as zooming and panning, are also not scene changes. Scene changes, that take several frames, such as wiping and dissolving, are not the target of this article.

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

Method of Detecting Scene Changes in Moving Pictures

       This article describes a method of detecting scene
changes in moving pictures. A scene change means a change of camera
position from one frame to the next. The inclusion of new objects or
the exclusion of old objects in a frame does not constitute a scene
change.  Camera movements, such as zooming and panning, are also not
scene changes.  Scene changes, that take several frames, such as
wiping and dissolving, are not the target of this article.

      Figure 1 shows the task flow for detecting scene changes.
1) One frame is inputted.  In this step, inputted frame is
transformed from analogue signals to digital data.
2) A Moving picture usually consists of color frame sequences.  The
transformed digital image may be in RGB, YIQ, or some other format.
In this article, the intensity of the frame is adopted as basic
information.  Thus, if the transformed image is in RGB style, its
intensity value is extracted from its RGB value.
3) It is not necessary to calculate all the pixels in the frame.
Experiments show that subsampling in the order of 1/100 is sufficient
for detecting scene changes.  Pixels are subsampled.  Random sampling
is generally the best method of subsampling, but rectangular sampling
is adequate for determining the overall characteristics of the frame.
4) The frame Difference Mean (FDM) between frame t and frame t-1 is
calculated by the following equation:

                ...