Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Interactive Method for Outlining Objects in a Displayed Image

IP.com Disclosure Number: IPCOM000114255D
Original Publication Date: 1994-Nov-01
Included in the Prior Art Database: 2005-Mar-28
Document File: 4 page(s) / 171K

Publishing Venue

IBM

Related People

Glasman, EH: AUTHOR [+2]

Abstract

Various applications require that objects or regions of interest in a digital image be outlined. Described is an interactive, semi-automatic method of doing this outlining. The user draws an approximate outline on a displayed color or grey level image (e.g., with mouse), and the outline is then refined to more nearly match the object border. The method is illustrated using the image in Fig. 1, a picture looking up towards the sun with several surrounding buildings. The sun is the object to be outlined.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Interactive Method for Outlining Objects in a Displayed Image

      Various applications require that objects or regions of
interest in a digital image be outlined.  Described is an
interactive, semi-automatic method of doing this outlining.  The user
draws an approximate outline on a displayed color or grey level image
(e.g., with mouse), and the outline is then refined to more nearly
match the object border.  The method is illustrated using the image
in Fig. 1, a picture looking up towards the sun with several
surrounding buildings.  The sun is the object to be outlined.

      STEP 1: Read the user-entered outline.  For this method, the
outline, which we call 0, is a simple, non-intersecting, 8-connected
curve.  The user outline is shown in Fig. 2.  In practice, it is
drawn by the user over the image.  For clarity, it is shown
separately here.  Also, in the Figure, it is intentionally drawn far
from the actual object boundary in places to illustrate how the
algorithm works when this happens.

      STEP 2: Morphically expand 0 d times, that is "fatten" 0 by
amount d on each side.  Call R the result of this expansion.  R is
(topologically) a donut-shaped region with a hole in its middle.
During the expansion, ensure that the hole in the center does not
disappear.  Fig. 3 shows R after the expansion, and Fig. 4 shows it
superimposed on the image.  Subsequent figures show the detail area
indicated approximately by the box in Fig. 4.

      STEP 3: Apply an edge detector to region R of the image I,
yielding a set of edges E.  For color images, the edge detector may
be applied to the luninance component computed from the R, G, and B
values; as a 3-D edge detector on (R,G,B); from one individual color
component; etc.  Fig. 5 shows computed edges in the area covered by
the bounding box of region R, and Fig. 6 shows the edges within R.
The boundary of region R is shown in grey.

      STEP 4: Extend the endpoints of each connected segment.  This
connects some of the disconnected edge segments.  The direction of
the extension is given by the tangent at the endpoint.

      STEP 5: Thin region R. The...