Motion video signal processing for recording or reproducing – Local trick play processing – With randomly accessible medium
Reexamination Certificate
2001-09-20
2003-07-01
Tran, Thai (Department: 2615)
Motion video signal processing for recording or reproducing
Local trick play processing
With randomly accessible medium
C386S349000, C348S552000, C707S793000
Reexamination Certificate
active
06587637
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a retrieving method and apparatuses therefor for video images on the air or video images in a data base or others and more particularly to a video image retrieving method and apparatuses therefor for performing high-speed retrieval by the help of features of video images.
2. Description of the Prior Art
Recently, multi-media information processing systems can store and indicate various types of information such as video and text to users. However, with respect to retrieval of them, a retrieving method using a language such as a keyword is mainly used. In this case, a keyword assigning operation is necessary and it is extremely expensive to assign a keyword to each frame of video having a large amount of information. Furthermore, since a keyword is freely assigned by a data base constructor, there is a problem imposed that when the viewpoint of a user is different from that of the data base constructor, the keyword will be useless. In these circumstances, a request for retrieval from a unique image feature in addition to the keyword is made. However, to retrieve information on the basis of the feature of an image, a high-speed comparison art between the video feature comprising enormous frames and the feature for the queried image is necessary. As a high-speed comparison art only applicable to video images, “Video retrieving method and apparatuses therefor” is proposed in Japanese Patent Application Laid-Open 7-114567. This method does not compare all the frames but compares only an image at the time of changing of cut of images so as to reduce the processing amount. By doing this, the high speed also suited to comparison of images on the air is realized. On the other hand, there is a problem imposed that a scene comprising only one cut or a scene in which the cut change timing varies with editing before or after cannot be compared satisfactorily. Furthermore, during retrieval, scenes other than the scene specified as a retrieval key are not searched in the same way as with other general data base systems, so that whenever scene retrieval becomes necessary, it is necessary to repeatedly compare a very large amount of video information from the beginning thereof to the last. The scene comparison process includes a number of processes such as processes to be performed commonly even if the scene to be retrieved is different as well as the feature extraction and reading processes and repetitive execution of such a process is of no use.
SUMMARY OF THE INVENTION
An object of the present invention is to solve the aforementioned problems and to provide an image retrieving method for comparing the feature of a target image to be retrieved and the feature of a sample image to be prepared for query at high speed without performing a keyword assigning operation for image retrieval and for detecting the same segment with the frame accuracy. A target image on the air or in the data base is applicable.
Another object of the present invention is to provide a method for detecting the same scene existing in the target image regardless of whether it is specified as a retrieval key beforehand in the same way at the same time with input of the target image.
Still another object of the present invention is to provide a video camera for comparing, when recording an image series inputted from moment to moment during picking up of images, those images with recorded images and recording them in association with matched images.
To accomplish the above objects, the present invention is a signal series retrieving method and apparatuses therefor in an information processing system comprising a time sequential signal input means, a time sequential signal process controller, and a storage, wherein the method and apparatuses sequentially input time sequential signals, sequentially extract features in each predetermined period of the inputted time sequential signals, convert the features sequentially extracted into a feature series corresponding to the inputted predetermined period series, compress the feature series in the direction of the time axis, store the compressed feature series in the storage, sequentially extract features from the time sequential signals to be retrieved in each predetermined period of the inputted time sequential signals, sequentially compare the features of the time sequential signals to be retrieved in each predetermined period with the stored compressed feature series, store the progress state of the comparison, and retrieve a signal series matching with the progress state from the time sequential signals to be retrieved on the basis of the comparison result between the stored progress state of the comparison and the features of the time sequential signals to be retrieved in each predetermined period.
More concretely, the present invention divides a video image to be compared into the segment-wise so that the feature of each frame is set in the variation width within the specific range respectively, extracts one or a plurality of features in each segment, stores it or them in correspondence with the address information indicating the position in the image in the segment, then sequentially inputs frame images one by one from video images to be retrieved, and when the feature series at an optional point of time in which the features of the frame images are sequentially arranged and the feature series in which the features in the segments constituting the stored images are sequentially arranged in each segment length have portions equal to or more than the specific length which can be decided to be mutually equivalent to each other, detects the portions as a same image. In this case, when they are equivalent to each other from the top of a segment, the present invention obtains the address information corresponding to the segment and when they are decided to be equivalent to each other from halfway of a segment, the present invention obtains the relative position from the top of the segment, and outputs a corrected value of the address information corresponding to the segment as a retrieval result. Furthermore, the present invention collects a frame image series inputted as a retrieval target in each segment so that the features of the frames are set in the variation width within the specific range, extracts one or a plurality of features in each segment, also stores the information corresponding to the address information indicating the position in the target image in the segment, and adds it to the target images to be compared next. Furthermore, with respect to the inputted feature series, when there are a plurality of video portions which are detected to be the same, the present invention groups them, associates them to each other, and stores them.
An apparatus realizing the aforementioned retrieving method comprises a means for dividing an optional image into the segment-wise so that the feature of each frame is set in the variation width within the specific range respectively, a means for extracting one or a plurality of features in each segment, a means for storing it or them in correspondence with the address information indicating the position in the image in the segment, a means for sequentially inputting frame images one by one from images to be retrieved, a means for retaining the feature series at an optional point of time in which the features of the frame images are sequentially arranged, a means for generating the feature series in which the features in the segments constituting the stored images are sequentially arranged in each segment length, and a means for deciding whether the feature series have portions equal to or more than the specific length which can be decided to be mutually equivalent to each other. The present invention also has a means for obtaining, when they are decided to be equivalent to each other from the top of a segment, the address information corresponding to the segment, when they are decided to be equivalent to each other from halfway of a segment, obtainin
Miyatake Takafumi
Nagasaka Akio
Mattingly Stanger & Malur, P.C.
Tran Thai
LandOfFree
Image retrieving method and apparatuses therefor 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 retrieving method and apparatuses therefor, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image retrieving method and apparatuses therefor will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3080610