Image analysis – Image compression or coding – Shape – icon – or feature-based compression
Reexamination Certificate
2000-11-17
2003-08-26
Couso, Jose L. (Department: 2621)
Image analysis
Image compression or coding
Shape, icon, or feature-based compression
Reexamination Certificate
active
06611628
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally relates to methods of image feature coding and image search and, more particularly, to a method of image feature coding and a method of image search in which features of analog or digital data of moving images or still images are extracted and coded so that the image data is searched using the coded feature.
2. Description of the Related Art
FIG. 1
is a block diagram showing a related-art image search system described in Denshi Jouhou Tsuushin Gakkai Rombunshi D-II Vol. 79-D-II No. 4, April, 1996, pp.476-483. Referring to
FIG. 1
, the system comprises a pre-processing unit
91
for segmenting a still image
201
, attaching a keyword to each of the resulting segments and outputting a conceptual keyword
203
and a scene descriptive keyword
204
; a search tool
92
for receiving a keyword
205
from a user
97
and retrieving the still image
201
by checking the input keyword against the conceptual keyword
203
and the scene descriptive keyword
204
.
The conceptual keyword
203
is a keyword indicating color information and feature of a segment. The scene keyword
204
is a keyword representing a feature of the segment using descriptive words relating to position, color, shape, size and orientation.
The pre-processing unit
91
of
FIG. 1
comprises a segmentation unit
93
for segmentation of the still image
201
; a conceptual keyword extraction unit
94
for extracting the conceptual keyword
203
from the color and feature of the segment produced by the segmentation unit
93
; and a scene descriptive keyword providing unit
95
for assigning the scene keyword
204
to the image feature of the segment produced by the segmentation unit
93
, by receiving a predicate description
202
selected by a user
96
from a set of predefined predicates.
The search tool
92
of
FIG. 1
comprises a feature identifying unit
98
for checking the keyword
205
selected by the user
97
against the conceptual keyword
203
and the scene descriptive keyword
204
from the pre-processing unit
91
.
A description will now be given of the operation.
When the still image
201
is supplied to the pre-processing unit
91
, the segmentation unit
93
segments the still image
201
. The conceptual keyword extracting unit
94
extracts the conceptual keyword
203
from the color and feature of the segment. More specifically, the conceptual keyword extracting unit
94
starts with a conceptual keyword associated with the color information to arrive at the conceptual keyword
203
.
The scene descriptive keyword providing unit
95
provides the scene keyword
204
to the image feature of the segment, by receiving the predicate description
202
from the user
96
.
When searching for the still image
201
, the user
97
inputs the keyword
205
, selected from a prepared set of conceptual keywords
203
and scene keywords
204
, to the search tool
92
. The feature identifying unit
98
retrieves the still image
201
requested by the-user
97
, based on the keyword
205
provided by the user
97
, the conceptual keyword
203
and the scene descriptive keyword
204
from the preprocessing unit
91
.
Since the target of the image search system described above is the still image
201
, there is a drawback in that it is difficult to search for moving images.
In addition, since not much consideration is given to how the keywords are provided and stored, a one-to-one correspondence between an image server and a client (search tool
92
) is a prerequisite. Therefore, according to the related art, an image search system where a large number of users are capable of searching for images using a variety of search tools via a network cannot be built.
SUMMARY OF THE INVENTION
Accordingly, a general object of the present invention is to provide a method of image feature coding and a method of image search in which the aforementioned drawbacks are eliminated.
Another and more specific object is to provide a method of image feature coding and a method of image search in which a large number of users can search for images using a variety of search tools.
The aforementioned objects can be achieved by an image feature coding method comprising the steps of: extracting segments of image areas from an image frame; attaching a segment number to each of the extracted segments; assigning a representative color to each of the extracted segments; computing a relative area of each of the segments with respect to the image frame; coding the representative color and the relative area to produce a feature of the image; and generating a feature stream corresponding to the image having the feature encoded therein.
The segments may be extracted from the image frame in accordance with color information, and the color information used in extracting the segments is assigned to the extracted segments as the representative color.
The segments from adjacent image frames may be checked for identity match, and those segments determined to match each other are given a same segment number.
The segments may be tracked from image frame to image frame so as to determine movement information relating to the segments that match each other.in identity, the movement information is generated by coding to produce the feature of the segments, and the feature stream, having the feature thus produced encoded therein, is generated.
An appropriate key frame that provides a key for a search is extracted from a group of image frames of a video signal, whereupon the segments are extracted from the extracted key frames.
A reduced image of the key frame may be generated by averaging pixels located in respective areas of the key frame, the reduced image is coded to produce the feature of the key frame, and the feature stream, having the feature thus produced encoded therein, is generated.
The aforementioned objects can also be achieved by an image searching method using first storage unit for storing image frames and a second storage unit for storing a feature stream having features of the image frames encoded therein, comprising the steps of: decoding the features stored in the second storage unit, in accordance with a search instruction from a user; and checking the decoded features against a search criteria provided by the user for identity match.
The features stored in the second storage unit may include a representative color of a segment constituting an area in the image frame, and the search criteria from the user may include the representative color.
The features stored in the second storage unit may include a relative area of a segment, constituting an area in the image frame, with respect to the image frame, and the search criteria from the user may include the relative area.
The features stored in the second storage unit may include movement information related movement between adjacent image frames, and the search criteria from the user may include the movement information.
The features stored in the second storage unit may include a reduced image of the image frame, the decoded feature may be checked against the search criteria from the user, and the reduced image may be presented to the user.
The features stored in the second storage unit may include information indicating whether a designated object is captured in the image frame.
The features stored in the second storage unit may include information indicating whether a designated object is captured subsequent image frames.
The features stored in the second storage unit may include information indicating whether a designated object is captured previous image frames.
Priority given to the decoded feature when checking the decoded feature against the search criteria from the user may be presented to the user.
A plurality of decoded features may be checked from an viewpoint against a plurality of search criteria from the user for a match from an overall perspective.
REFERENCES:
patent: 5619338 (1997-04-01), Nakai
patent: 5771334 (1998-06-01), Yamauchi et al.
patent: 5867221 (1999-02-01), Pullen et al.
patent: 5
Asai Kohtaro
Chow James
Sekiguchi Shun-ichi
Yamada Yoshihisa
Birch & Stewart Kolasch & Birch, LLP
Couso Jose L.
Mitsubishi Denki & Kabushiki Kaisha
LandOfFree
Method of image feature coding and method of image search does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of image feature coding and method of image search, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of image feature coding and method of image search will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3121891