Image analysis – Color image processing – Pattern recognition or classification using color
Reexamination Certificate
2000-02-22
2003-09-16
Chang, Jon (Department: 2623)
Image analysis
Color image processing
Pattern recognition or classification using color
C382S199000
Reexamination Certificate
active
06621924
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a contour extraction apparatus and a method thereof for extracting the contour of an aimed target object from an image, and a program recording medium which records a program adopting the contour extraction method.
When extracting an image of an aimed object from an image and synthesizing the extracted object image with another image, a soft key is used. A soft key generator which generates a soft key is normally constructed by a closed contour curve generator section, a closed end curve generator section, and a soft key generator section. The closed end curve generator section generates closed end curves of both ends from a closed contour curve generated by the closed contour curve generator section, and the soft key generator section uses the closed end curves to generate the soft key.
Of these sections, in the closed contour curve generator section, when a user uses a two-dimensional pointing device such as a mouse or a tablet not shown as an input means and specifies a representative point on an edge of an image, a color space projection axis which maximizes the contrast in the periphery of the edge within an update sector of a closed contour curve is calculated from the coordinates of the specified representative point on the edge, and a new contour curve within the update sector is generated based on the color projection axis or the like. A new closed contour curve is generated from this new contour curve.
Also, in the closed contour curve generator section, more precise area extraction is enabled through input information from an operator.
At first, a conventional method for generating the color projection axis will be specifically explained with reference to FIG.
1
.
It is generally known that the color distribution Cf of the foreground area and the color distribution Cb of the background area at the periphery of an edge have a banana shape extending in the brightness axis direction. Therefore, conventionally, the Axis
1
in the brightness direction shown in
FIG. 1
is taken as the color projection axis.
Next, explanation will be made of a conventional method which performs more precise area extraction by inserting input information from an operator.
The first conventional technique including a closed contour curve generator section with inserted instruction information is of a type in which an operator determines information that has been automatically extracted but is unnecessary and information that could not be automatically extracted but is necessary, with respect to contour information automatically extracted by any processing, and deletes or adds those information items. Specific examples of prior art belongs to this type will be “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 3-291770) and “Image Edit Apparatus” (Japanese Patent Application KOKAI Publication No. 1-175076).
Further, the second conventional technique is of a type that an operator determines an insufficient portion of an initial input with respect to input information for performing contour extraction processing and further provides a more accurate input to make calculation again. Specific examples of prior art which belong to this type will be “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 4-152481), “Pattern Mask Preparation Method” (Japanese Patent Application KOKAI Publication No. 1-1554
6
), “Clipping Mask Preparation Method” (Japanese Patent Application KOKAI Publication No. 4-68763), and “Intelligent Scissors for Image Composition” (Eric N. Mortensen and William A. Barrett, Computer Graphics Proceeding, Annual Conference Series, 1995, ACM SIGGRAPH, pp. 191-198).
Brief explanation will now be made of operation of each of the conventional techniques described above which include a closed contour curve generator section with inserted input information from an operation.
In the “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 3-291770), an edge detection operator is used to extract a contour in the first automatic extraction, and an operator specifies both ends of a contour which could not be extracted by edge detection processing, thereby to add a line segment connecting the both ends as a contour.
The “Intelligent Scissors for Image Composition” create several relay points on a contour and operates to search and extract a contour point between the points. This search method is based on a search rule using edge information of an image and searches such a path which connects the relay points so as to pass above the edge if possible. It is possible to extract automatically a considerably complicated contour shape. Therefore, this method is advantageous in that a contour can be extracted with use of a relatively small number of relay points.
The “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 4-152481) also creates several relay points on a contour and operates so as to search and extract a contour point between the relay points. This method operates to apply a bias in the search direction depending on the position of the relay point when searching a contour. Therefore, a contour extending in a desired direction can be easily extracted by changing relay points.
In the first conventional technique described above, the accuracy of the final contour greatly depends on the accuracy of first automatic extraction. A problem hence arises in that it cannot be much expected that the contour can be more accurately corrected by later correction operation by an operator.
For example, in the “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 3-291770), manual addition of a contour can be achieved in form of a line segment, so there is a problem concerning the accuracy of the shape.
Also, in the “Image Edit Apparatus” (Japanese Patent Application KOKAI Publication No. 1-175076), the areas divided at first should not be too small in order that correction operation is not complicated. Therefore, color quantization need to be rough appropriately. If color quantization is rough, however, there is a problem that the shape of the divided area and the true object contour shape easily disagree with each other so that the accuracy of the automatically extracted contour shape is lost.
Meanwhile, the second conventional technique is advantageous in the accuracy of the contour because a more accurate input is provided and recalculation is performed on an area where a sufficient result could not be obtained from automatic extraction. However, this second conventional technique includes the following problem.
In the “Pattern Mask Preparation Method” (Japanese Patent Application KOKAI Publication No. 1-155476) and the “Clipping Mask Preparation Method” (Japanese Patent Application KOKAI Publication No. 4-68763), operator's correction functions to change partially an input concerning the background color or foreground color as described above and perform recalculation as to contour extraction in the corresponding portion thereby to correct the contour shape. However, the color input operation is equivalent to operation of carrying out level adjustments of keys in a chroma key device, as it were. Since this kind of operation does not achieve direct inputting of information concerning a contour shape, it is difficult for an operator to carry out intuitively correct inputting, and therefore, try and error are repeated several times to attain a desired contour shape.
In contrast, in the “Image Contour Detection Method” (Japanese Patent Application KOKAI Publication No. 4-152481) and the “Intelligent Scissors for Image Composition” (which will be hereinafter called a conventional technique for inputting a relay point), operator's correction is to operate a relay point on a contour and functions to make recalculation as to a contour shape concerning a corrected relay point. In this method, an input concerning a contour shape is directly supplied so that it is relatively easy for the o
Ishii Takahiro
Ogino Tomotaka
Totsuka Takashi
Chang Jon
Le Brian
Sonnenschein Nath & Rosenthal LLP
LandOfFree
Contour extraction apparatus, a method thereof, and a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Contour extraction apparatus, a method thereof, and a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Contour extraction apparatus, a method thereof, and a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3009582