Image analysis – Image segmentation
Reexamination Certificate
1999-09-10
2002-09-03
Au, Amelia M. (Department: 2623)
Image analysis
Image segmentation
C382S199000, C382S203000, C358S453000
Reexamination Certificate
active
06445819
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image processing method, an image processing device, and a recording medium, and in particular, to an image processing method for extracting a region which is assumed to be a region corresponding to a main portion in an image, to an image processing device to which the image processing method may be applied, and to a recording medium on which is recorded a program for executing the image processing method at a computer.
2. Description of the Related Art
When viewing a photograph of a person, the region to which the most attention is paid is the face of the person. For example, in a case in which an original image recorded on a photographic film is exposed and recorded (recorded by plane exposure or scanning exposure) onto a recording material such as a photographic printing paper, it is desirable to control the exposure such that the color and density of a person's face are appropriate. However, to realize such control of exposure, the color tint and the density of the region corresponding to the person's face in the original image must be accurately sensed. Further, among various image processings which have been developed in order to improve the image quality of an image expressed by image data obtained by the reading of the image, there are special image processings (such as correction of the density only at certain portions or red-eye correction or the like) which are carried out only on a region corresponding to a person's face within the image or only on a portion of such a region. In order to carry out such processings, the position and size of the region corresponding to a person's face within the image must be sensed accurately.
There have been proposed various conventional techniques for extracting a region assumed to correspond to a main portion such as a person's face within an image. For example, Japanese Patent Application Laid-Open (JP-A) No. 8-184925 discloses a face region extraction method in which, on the basis of image data, one configuration pattern of configuration patterns peculiar to portions of a person existing in an image (e.g., configuration patterns expressing the contour of a head portion, the contour of a face, the inner structure of a face, the contour of a torso, and the like) is searched for. In accordance with the size and orientation of a detected configuration pattern and the positional relationship between the face of a person and a predetermined portion of the person expressed by the detected configuration pattern, a region assumed to correspond to a person's face is set, other configuration patterns different from the detected configuration pattern are searched for, and a region (face region), which is within the set region and confirmed as being appropriate as a person's face and is assumed to correspond to a person's face, is extracted.
However, among the various techniques which have been proposed, none can always precisely extract a region corresponding to a main portion such as a person's face in an image, and there are cases when mistakes are made in extracting a region corresponding to a main region. When such mistakes are made, a drawback arises in that the results of image processing or exposure control carried out by using the results of extraction are inappropriate.
For example, in the case of an image in which a person is photographed by using a flash in a back-lit setting, if a high-brightness background portion within the image is erroneously extracted as a region corresponding to a person's face, when the amount of exposure at the time of exposing and recording the image onto a recording material is controlled on the basis of the color and the density of the extracted region, the recorded image will be of poor image quality as the region corresponding to the person's face will be colored-in black. Further, even in cases in which a special image processing is to be carried out only on the extracted region or a portion thereof, if the special processing is carried out only on a region which is different than a region which corresponds to a person's face within the image, the image quality of the image expressed by the image data deteriorates.
SUMMARY OF THE INVENTION
The present invention was developed in light of the aforementioned circumstances, and an object thereof is to provide an image processing method, an image processing device and a recording medium in which effects due to errors in extracting a region assumed to correspond to a main portion within an image can be mitigated.
The present inventors selected an arbitrary extraction method from among various extraction methods for extracting a region assumed to correspond to a main portion such as a person's face in an image. They repeatedly conducted experiments of extracting regions corresponding to main portions from a large number of images by using this selected extraction method, by varying the processing conditions of the extraction method (e.g., the threshold value for determining whether a region is a main portion) and by varying the extraction method itself. By comparing the regions extracted by these respective extractions, the present inventors discovered that, in most cases, the extracted regions which corresponded to or substantially corresponded to a main portion within the image overlapped other extracted regions and the degree of overlapping with other extracted regions was high. Thus, the present inventors hypothesized that the probability of an extracted region being a region corresponding to a main portion within an image is strongly correlated with the degree of overlapping with other extracted regions.
Among the various extraction methods, due to differences in the algorithms thereof, there are extraction methods in which there is no overlapping of extracted regions even in a case in which plural regions are extracted in a single extraction process, and there are extraction methods in which there is the possibility of overlapping of extracted regions occurring when plural regions are extracted in a single extraction process. (For example, in an algorithm which traces an edge in an image and extracts a region corresponding to a main portion, if edge tracing is carried out by using respectively different points as the starting points, different regions in which portions thereof overlap may be extracted as regions corresponding to main portions.) The present inventors confirmed that, even in the latter type of extraction methods, it is often the case that plural regions are extracted by a single extraction process, and extracted regions which overlap with another extracted region correspond or substantially correspond to a main region within the image, and, as mentioned above, the probability of an extracted region being a region corresponding to a main portion in an image is strongly correlated to the degree of overlapping with another extracted region.
On the basis of the above, the image processing method of the first aspect of the invention comprises the steps of: on the basis of image data, extracting candidate regions which are assumed to correspond to main portions in an image represented by the image data; determining, for each of candidate regions which overlap with another candidate region among the extracted candidate regions, a degree of overlapping with another candidate region; and for an extracted candidate region, evaluating the probability of being a region corresponding to a main portion in the image, in such a manner that, for a candidate region overlapping with another candidate region, an evaluation of the probability of being a region corresponding to a main portion within the image is higher than said evaluation for a candidate region not overlapping with another candidate region, and becomes higher as a degree of overlapping with another candidate region increases.
In the first aspect of the present invention, first, on the basis of the image data, a candidate region assume
Au Amelia M.
Dastouri Mehrdad
Fuji Photo Film Co. , Ltd.
LandOfFree
Image processing method, image processing device, and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing method, image processing device, and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing method, image processing device, and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2856832