Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1998-07-02
2002-04-16
Kelley, Chris (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C375S240000, C382S176000, C382S170000, C382S171000, C382S270000
Reexamination Certificate
active
06373891
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention pertains to a system and method for recognizing unique image sequences, commercials, for instance, in which characteristics are extracted from the individual images regarding their brightness, digitized and compared to a reference sample.
Image sequences that consist of a number of successive images tied together in terms of their contents are to be recognized in, for instance, television at each new broadcast. These image sequences can be commercials, old films or video clips, or political items such as campaign segments. Common to all the aforesaid applications is the fact that the dissemination of the image sequences is to be recorded for legal reasons due to fees to be collected or for statistical reasons. In this regard, unique image sequences means that, in their content, the images remain preserved in their original form, that is to say, all image points are unique in brightness and hue and thus unchangeable. In copying by high-quality systems, this unique arrangement is not changed or supplemented in any value. The concept of “uniqueness” thus also refers to copied image sequences.
Commercials are typically broadcast by television stations at times that are particularly interesting to the advertisers, because the reach is particularly large. There is thus an interest for the advertiser to check whether its commercial was actually broadcast at the selected time. A commercial changes appearance over time, i.e., it becomes shorter or some images are changed or completely replaced. This new image sequence is now to be differentiated from the original version.
A method for identifying unique image sequences is known from DE 43 09 957 Cl. Therein, individual image elements, known as pixels, are gathered into groups or blocks, which are called clusters and whose luminance or brightness values are digitized and compared as characteristics with the characteristics of known images or reference samples.
Because of the necessary data compression or data reduction, the probability of false recognition of an image is relatively high. In order to reduce the number of false recognitions, several images in sequence are therefore extracted in the known method. The disadvantage here is that successive images are normally very similar so that the characteristics found frequently lead to chance similarities, without similar images actually being present. These so-called chance false recognitions do not in themselves impair the recognition of a commercial, but do lead to a large accumulation of data, which burdens the computer.
Therefore, a method of preventing the accumulation of data and reducing the number of false recognitions is desired.
A device for carrying out a method for identifying unique image sequences is also known from DE 4309 957 Cl. The known device has the drawback that an extraction of characteristics is possible only from successive images, so that the characteristics extracted from the images correlate with one another and lead to a high number of false recognitions.
Therefore, a device with which a quasi-stochastic extraction of characteristics is possible is also desired.
SUMMARY OF THE INVENTION
The problem of relatively large data accumulation and false recognitions is solved according to the invention in that the characteristics are decorrelated by a quasi-stochastic method across several images.
By virtue of the fact that, instead of extracting characteristics from adjacent images, characteristics are extracted according to a quasi-stochastic method across several images, the chance similarities of successive images can be neglected, since the connections of successive images are broken up. The number of chance false recognitions and the associated accumulation of data burdening the computer are thus reduced.
According to a preferred embodiment of the invention, selected characteristics of individual images are written in an ordered sequence into a memory, organized as a shift register, of an image carousel and read out according to a quasi-stochastic process in a nonselective access. The access to the characteristics is done here with a maximum jump width across the images. At the same time, access to the characteristics is done with the greatest possible distance inside the images. By virtue of this measure, a decorrelation of the characteristics is achieved, because the temporal and spatial connections are broken up.
According to another preferred embodiment of the invention, a two-stage processing of the method is conducted, with a first stage for general recognition and a second stage for detailed recognition. General recognition is so reduced in data accumulation that it is suitable for real-time. It permits the recognition of a commercial without having to be able to determine any mutations which might be present. Detailed recognition takes place only if the general recognition process has assigned an image sequence or commercial to the existing reference example. Detailed recognition then provides the final certainty and reveals spatial and temporal mutations.
According to an additional preferred embodiment of the invention, the change of brightness inside a cluster formed from spatially cohesive pixels is utilized.
In the process, the clusters are subjected to a discrete cosine transform. One of the low-frequency alternating coefficients is reduced to its sign and used as the characteristic. An image sequence is divided up into time slices of constant length, each time slice representing an autonomous unit. which is correlated. The images of a time slice are thus coded in an extremely data-saving manner. The recognition or general recognition of a commercial results if the individual time slices are recognized in the correct sequence and in the right spatial intervals.
The invention further pertains to a system for carrying out the method.
The problem of quasi-stochastic image extraction of characteristics is solved in that an image carousel having a memory organized as a shift register is connected at one end via a DCT transformer to a video decoder with raster/cluster converter and at the other end via a correlator to a reference memory such that an image sequence fed by the receiver into the video decoder can be supplied to the correlator as a characteristics vector and compared to a reference sample that can be stored in the reference memory.
A quasi-stochastic extraction of characteristics is enabled in a simple manner by the use of the image carousel, wherein the characteristics of individual images are decorrelated.
According to another preferred embodiment of the invention, a branch is provided between the DCT transformer and the image carousel to a detailed recognition memory, which is constructed as a FIFO (first-in, first-out) shift register in which all the characteristics of an image sequence can be stored.
By the use of the FIFO shift register, it is possible to carry out detailed recognition only after the general recognition, which has taken place in real-time.
Additional details of the invention are seen from the extensive description below and the attached drawings, in which preferred embodiments of the invention are illustrated as examples.
Other objects and features will be in part apparent and in part pointed out hereinafter.
REFERENCES:
patent: 5559557 (1996-09-01), Kato
patent: 5742704 (1998-04-01), Suzuki et al.
patent: 43 09 957 (1994-07-01), None
An Shawn S.
Kelley Chris
Senniger Powers Leavitt & Roedel
LandOfFree
System and method for image sequence processing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for image sequence processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for image sequence processing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2878958