Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
2000-09-20
2004-09-07
Mariam, Daniel (Department: 2621)
Image analysis
Pattern recognition
Feature extraction
Reexamination Certificate
active
06788816
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image processing method and apparatus, and more particularly to a method of and an apparatus for extracting, from image data of a picture image including an image of a body which makes an object of contour extraction, a contour of the object body based on attributes of individual points of the picture image. More specifically, the present invention relates to a contour extraction method and apparatus suitably applied to an apparatus which extracts a particular area such as an internal organ or a tumor from image data of a picture image picked up by an MRI apparatus or a CT scanning apparatus.
2. Description of the Related Art
Various proposals have conventionally been made to extract an image of a particular body from a given picture image. For example, a contour extraction apparatus is disclosed in Japanese Patent Laid-Open No. 299366/1998 (hereinafter referred to as “document 1”) which extracts, based on image data representative of a tomographic image of an internal organ from an ultrasonic diagnosis system or an MRI picture image, a region of the internal organ or a region of a tissue. In the contour extraction apparatus, in order to allow a region of an appropriate size to be extracted using a suitable threshold value, when a finite difference between time series data representative of an area within an extracted contour and smoothed data of the time series data of a plurality of tomographic images obtained at predetermined frame time intervals exceeds a predetermined threshold value, the threshold value for the contour extraction is varied and contour extraction is preformed again with the varied threshold value.
In the apparatus disclosed in the document 1 mentioned above, whether each point of a picture image of image data is an internal point or an external point of an area is first determined using a predetermined first threshold value based on a criterion of whether or not the value of the point exceeds the threshold value. Then, the area or the volume of the region determined by the method described above is determined from the picture images at the different times, and the time series data of the area or the volume of the region are smoothed, whereafter it is discriminated whether or not a finite difference between the data before smoothing and the data after smoothing exceeds a second threshold value.
If it is discriminated that the finite difference exceeds the second threshold value, the first threshold value is varied and the contour extraction performed first is performed again with the varied first threshold value.
The conventional contour extraction apparatus disclosed in the document 1 mentioned above, however, has the following problems.
First, while the apparatus disclosed in the document 1 uses two different threshold values in order to extract a region, a detailed method for determining the threshold values is not disclosed in the document 1, and each time data which make an object of region extraction changes, the threshold values must be determined by a trial-and-error scheme.
However, for example, with regard to an MRI image, since the appropriate threshold values vary depending upon of which part of the human body the tomographic picture image is, the apparatus disclosed in the document 1 has a problem that, if an inappropriate threshold value is used, then the contour of an internal organ different from an intended internal organ is extracted in error.
Second, the apparatus disclosed in the document 1 has a problem that, since it is necessary to use time series data of picture image data, a large amount of data is required in order to determine one region extraction picture image.
Third, in the apparatus disclosed in the document 1, a finite difference between data before smoothing and data after smoothing is calculated, and when the finite difference exceeds the second threshold value, it is necessary to vary the first threshold value and repeat the same procedure with the varied first threshold value. Therefore, the apparatus disclosed in the document 1 has a problem that a large amount of calculation time is required.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a contour extraction method and apparatus by which a contour can be extracted at a high speed with a high degree of accuracy without the necessity for setting a threshold value explicitly.
In order to attain the object described above, according to the present invention, from picture image data of a picture image including an image of a body which makes an object of contour extraction, region belonging probabilities with which individual points of the picture image belong to regions are calculated not based on values themselves of the image data but based on attributes of the points of the picture image, and the regions to which the individual points of the picture image belong are delimited using the region belonging probabilities, and then a boundary between the regions is extracted as a contour.
More particularly, according to an aspect of the present invention, there is provided a contour extraction method for delimiting, from picture image data of a picture image including an image of a body which makes an object of contour extraction, regions to which individual points of the picture image belong based on attributes of the points of the picture image and extracting a boundary between the regions as a contour, comprising a first step of initializing parameters which define a mixed probability distributions of the attributes of the points of the picture image, a second step of calculating region belonging probabilities with which the points of the picture image belong individually to the regions, a third step of updating the parameter so that the mixed probability distribution may be increased, a fourth step of calculating an evaluation function to be used as a scale for favorableness of estimation from the mixed probability distribution defined by the updated parameters, a fifth step of delimiting the regions to which the points of the picture image belong based on the region belonging probabilities, a sixth step of extracting a boundary between the delimited regions, and a seventh step of discriminating based on the region belonging probabilities whether or not the points of the picture image are region internal points which are points within a determined one of the regions or region external points which are points outside the determined region.
Preferably, the second, third and fourth steps are repetitively performed until a condition determined in advance is satisfied.
The fifth step of delimiting the regions to which the points of the picture image belong may include the steps of determining one region internal point and setting the region internal point as an initial set to a region internal point set, acquiring neighboring points to the point belonging to the region internal point set and setting the neighboring points as an initial set to a boundary candidate set, selecting one of the points of the boundary candidate set which belongs to the region internal point set and adding the selected point to the region internal point set, sending, at a point of time at which there remains no point to be newly added to the region internal point set any more, the region internal point set to the sixth step, acquiring, when the region internal point set is to be selected, neighboring points to each of the points belonging to the boundary candidate set and adding the points which belong to the boundary candidate set to the region internal point set if all of the neighboring points are region internal points, adding, if the neighboring points include at least one region external point, the point or points which belong to the boundary candidate set to a boundary point set, and adding one or those of the region internal points belonging to the neighboring points which are not added to the region internal point set to the boundary candidate set.
The sixth step of extrac
Dickstein Shapiro Morin & Oshinsky LLP.
Mariam Daniel
NEC Corporation
LandOfFree
Contour extraction method and apparatus 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 method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Contour extraction method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3238929