Browse Prior Art Database

Detecting the Boundary of a Moving Object on a Motion Picture

IP.com Disclosure Number: IPCOM000078868D
Original Publication Date: 1973-Mar-01
Included in the Prior Art Database: 2005-Feb-26
Document File: 5 page(s) / 58K

Publishing Venue

IBM

Related People

Kaneko, T: AUTHOR [+2]

Abstract

The detecting of a boundary of a significant object in a motion picture is important in many image recognition problems. In the publication of A. Rosenfeld, "Picture Processing by Computer", Academic Press, New York, pp. 132-133, there are described several methods for such boundary detection. However, the methods described in this publication are essentially applicable to individual pictures, but not to those pictures which constitute a motion picture. In this latter connection, it is realized that the consecutive pictures of a motion picture are highly correlated, and a boundary of an object on the present frame changes in shape only very slightly from the immediately previously occurring frame. There is described herein, an efficient method for detecting such boundary of a moving object on a motion picture.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 30% of the total text.

Page 1 of 5

Detecting the Boundary of a Moving Object on a Motion Picture

The detecting of a boundary of a significant object in a motion picture is important in many image recognition problems. In the publication of A. Rosenfeld, "Picture Processing by Computer", Academic Press, New York, pp. 132-133, there are described several methods for such boundary detection. However, the methods described in this publication are essentially applicable to individual pictures, but not to those pictures which constitute a motion picture. In this latter connection, it is realized that the consecutive pictures of a motion picture are highly correlated, and a boundary of an object on the present frame changes in shape only very slightly from the immediately previously occurring frame. There is described herein, an efficient method for detecting such boundary of a moving object on a motion picture.

Reference is made to Fig. 1, wherein there is shown a block diagram of apparatus which may be suitably employed to carry out the method. In the operation of this apparatus, a scanner 12 scans each frame of a motion picture 10 and produces in response thereto a video signal. It is of course realized that if the picture is recorded on video tape, then the scanning step can be eliminated, the video signal being produced at the output of a tape recorder 14. Either the video signal output of scanner 12 or video tape recorder 14 is applied to a quantizer 16 wherein the video signal is quantized, the output of quantizer 16 being applied to a digital computer 18. The magnetic disk unit 20 may be supplied in the event that the memory capacity of the computer 18 is limited, Whereby a large number of digitized images can be stored in this unit. The digitized image can then be displayed on a suitable display unit 22, and an interactive device such as a light pen 24 may be advantageously supplied to interact with the display unit 22 and digital computer 18, whereby the X-Y coordinates of a point on a picture can be manually identified.

In considering the described method, it is assumed that the boundary of the moving object Which it is desired to detect is already known and available. It can be detected manually through the use of the light pen 24. Alternatively, it may be detected by a more sophisticated automatic method, such as is described in the disclosure of C. K. Chow and T. Kaneko, "Boundary Detection Method and Algorithm," IBM Technical Disclosure Bulletin, Vol. 14, No. 4, September 1971, pages 1301 - 1304.

It is readily appreciated that a boundary is well-defined on a picture such as a binary image with a smooth edge. However, on a gray level picture containing noise it is quite difficult to define, its location often being subjective and depending upon its content. The described technique is based upon the concept that a boundary is a smooth continuous curve, stretching over points with large- intensity changes (or high-spatial derivations). Each portion of a bounda...