The present invention relates to a method for facilitating the archiving of video content.
The so called term “video clip” appearing in this specification means a finite duration video content along with associated audio tracks, whether in digital or analog formats. Since video content consists of time-consecutive frames of video images. A video clip consists of a finite number of time-consecutive number of video images, along with the associated audio tracks of the same duration.
The so called term “fingerprint” appearing in this specification means a series of dot information, in which each dot information is selected from a frame of pattern of television signals, and a plurality of frames can be selected from the television signals, and one or more dot data can be selected from one frame of pattern of television signals, so that the so called “fingerprint” can be used to uniquely identify said television signals.
The so called term “visually identical” appearing in this specification means that two video content segment are visually identical if they are obtained from a single video image capture or recording device at the same time. In other words, they originate from a single video source and at the same time, i.e., a single time-space video source. For example, two copies of a single video tape are visually identical because they are from the same source. Two versions of compressed video data streams are visually identical if they are encoded and/or re-encoded from the same video content source, despite the fact that they may have different compression formats, bit rates or resolutions. On the other hand, two different video recordings of the same scene, but shot from two different cameras, or two different video recordings of the same scene but shot at different times from the same camera, are NOT visually identical because they are not created from a single time-space video source. When there are two different video recordings of different durations, a section of each recording may still be visually identical.
The fingerprint is used to seek out visually identical video segments between two different video content pieces. The content pieces may be in analog recording format, or in digital compressed format, or in digital uncompressed format. When two video clips are to be compared to determine if they are visually identical, in stead of comparing the two video contents, an automatic procedure can be deployed to compare the fingerprints obtained from each of the video clips. If the fingerprints match each other, then it is to say that the video clips are visually identical to each other.
A typical application is to use the technique to perform fingerprint based identification of video content. Specifically, known video clips are first registered into a fingerprint database, and when new video clips are obtained, their fingerprints are compared with the fingerprints already in the database to determine if the new content is visually identical to a previously registered video clip.
In this document, the terms “video”, “video content”, and “video signals” generically represent the same concept, i.e., visual information that can be displayed on television or computer monitors. The terms “video frames”, “digital video images”, and “video image frames” generically represent digitized video images, i.e., time-consecutive images that together form the motion video content. The video images, as part of the same video content, have the same number of video samples formed in rows and columns. The number of samples in a row is the width or horizontal resolution of the image, and the number of samples in a column is the height or the vertical resolution of the image.
In addition, the term “fingerprint” or “fingerprint data” represent the data formed by sampling consecutive video frames. The fingerprint or fingerprint data can be used to determine if two video contents are visually identical or not. Continuous samples of video frames form fingerprint data streams, or fingerprint streams. To better organize the fingerprint stream, sometimes, it is necessary to partition a continuous fingerprint stream into multiple segments. These segments are so called “fingerprint data segments” or just “fingerprint segments”.
With digital compression technology, it is possible to carry many television channels in the spectrum space of a single analog channel. Because of this, more television channels are transmitted over the same spectrum, and there are more channels for viewers to watch. Digitally compressed video and audio signals are binary data streams that can be transmitted, stored and played out as computer data files or streams. Therefore, digital video/audio data are typically handled in digital forms during production, transmission and storage phases.
Most of the video content in television broadcast, cable broadcast and on the interne originates in digital formats. Most of the content is already produced and stored in digital storage devices before it is distributed to consumers. For the purpose of managing advertising activities and marketing campaigns, government regulation enforcement, market research, and broadcasting signal monitoring, there is a need to continuously monitor the video signals as they are distributed to viewers.
In prior art, the operator may be given a video clip and be asked to search through the archived recordings to see where and when the video clip has shown up in video distributions in the past. In other words, the operator may be asked to search through the archived recordings to seek video content that is visually identical to the given video clip. For example, advertisers may want to determine if a particular commercial video has been distributed properly over the last year in certain geographic areas, so that they can track the effectiveness of their advertising campaign.
There are several problems with the above. The first problem is the fact that video content typically consumes massive amount of storage capacity. For example, a single channel of television content broadcast for 24 hours would consume at least 10 GB. For the purpose of monitoring thousands of television channels for a year or more, the storage capacity requirement can easily get into the Petabyte (1 PB=1,000,000,000,000,000 bytes) ranges.
The other problem is that even if the content is saved in storage systems, the cost to maintain, search and archive the content data can be too expensive for most users. This cost is reflected in terms of storage, computation and network hardware systems needed to perform the tasks. In many scenarios, the information to be archived and searched is not on the specific content itself, but more specifically related to when and where the content is distributed. This information can be valuable for content owners, marketers and relevant government regulators to track the coverage of specific video content. The purpose of this invention is to provide a method to facilitate this capability.
Therefore, there is a need to provide a method for facilitating the archiving of video content without a huge storage capacity required, and to be able to search the information easily at low hardware cost. There is also a need to collect statistics and extraction additional information from the archived video information automatically.
It is object of the present invention to provide a method for facilitating the archiving of video content, which can collect statistics and extraction additional information from the archived video information automatically.
It is another object of the present invention to provide a method for facilitating the archiving of video content, which can extract fingerprint information from video content for the purpose of archiving without the huge storage capacity required.
It is another object of the present invention to provide a method for facilitating the archiving of video content, which can search through video fingerprint data for identifying information and statistics about historical recording of video content easily at low hardware cost.
It is another object of the present invention to provide a method for facilitating the archiving of video content, which can extract fingerprint information from video content across geographically different remote locations and over long period of time for the purpose of archiving without the huge storage capacity required.
Therefore, according to the present invention, there is provided a method for facilitating the archiving of video content, wherein said method at least comprises distributing collectors at which video signals are distributed, the video signals being distributed in many geographically different places, and/or over different time periods; making said video signals go through a fingerprint extraction process to form extracted fingerprint data via collectors; and sending the fingerprint data collected from the collectors to a data center for archiving via a data path.
Preferably, the data path between the collectors and the data center is a network based on the IP protocols, wireless networks, telephone modem and telephone networks, or removable storage devices hand carried physically.
Preferably, the video signals that the collectors receive are either in analog or digital format.
Preferably, during the fingerprint extraction process, the incoming video frame data is first stored into a frame buffer, then sub-sampled by a sub-sampler to obtain extracted fingerprint data stream, a divider is used to break the sub-sampled fingerprint data stream down into video fingerprint data segments along the video frame boundaries, the video fingerprint data segments is then combined by a formatter with additional tracking information to form data packets which are then sent out via the transfer buffer.
Preferably, each of the data packets is self contained with its own data header, location, channel, segment number, time-stamp and other auxiliary information, the fingerprint data is also part of the packet, the data header contains information as head flags, packet length information, the number of video frames associated with the samples in this packet, and the manner at which the samples are made.
Preferably, the data packets are transferred to a data center and become part of the fingerprint archive database, and each packet appears as an entry in the database.
Preferably, said additional information at least comprises location, time, channel, and/or network origination address.
Preferably, said fingerprint data is extracted from video signals captured by remote collectors.
Preferably, the collector comprises an analog to digital (A/D) converter and a fingerprint extractor, the analog to digital (A/D) converter is used to first digitize the analog video signal into digital video sample images before they are sent to the fingerprint extractor.
Preferably, the collector comprises a receiver converter and a fingerprint extractor, the video signals in digital compressed video data format first go through the receiver converter to perform decompression of the video signals and deliver the digital decompressed video signals to the fingerprint extractor.
By using fingerprint in stead of the video content itself, it is possible to archive huge amount of video content without the associated storage capacity required. For example, it is possible to maintain a fingerprint archive for 1000 television channels, on a 24 hours basis, for 10 years, with a fingerprint archive database of no more than 40 terabytes (1 TB=1,000,000,000,000 bytes). However, to store video of the same duration at 1 million bits per second, the storage required would be 40 petabytes (1 PB=1000 TB), which is 1000 times higher than maintaining a fingerprint archive.
In addition, the fingerprint archive can be automatically compared for searching applications. For example, a user may have a video clip of 15 seconds, and wants to know if this same video clip has ever appeared in the past 10 years in any of the 1000 television channels. By using the technique from this invention, it would be possible to obtain information on: time of its appearance; location of the distribution (where the video content was originally recorded); and channel number, i.e., which television channel actually carried this signal, so that it is possible to know whether the video has appeared anywhere in the recorded archive.
By using the invention described in this document, it's possible to perform a search like the above across a huge archive which would be impossible if the video content itself, instead of the fingerprint data, is used.
In the present invention, discussions are focused on the handling and processing of digital video data streams. The method can be extended to audio data streams by using variations of the techniques and will not be discussed here.
Video monitoring requires that video content distribution activity be identified by time, location, channel assignment or network origination address. In addition, the monitoring must be identified by its source content. The preferred way to monitor video content is to have a recording of the video content as it's distributed. Such recording typically is in digital formats, stored as data files in a computer system. The recording also has the additional information attached to the recording itself, such as time, location, etc. At a later time, when an operator decides to verify the video content distribution, he or she can simply retrieve that video recording and view it in person.
In this invention, we provide a method for facilitating the archiving of video content.
As shown in
Alternatively, the extracted fingerprint data can be stored in a local storage 25 that is not removable but can be transferred via the network interface 23 and the network connection to the data center 4 at non-regular time intervals or at pre-scheduled time. This is shown in
The data organization within the transfer buffer 205 is as shown in
The data packets 300 are transferred to the data center 4 and become part of the fingerprint archive database. The database will be organized by the data packets 300 received, as shown in
The fingerprint archive database can hold such packets 300 from potentially large number of collectors 2 over a long period of time durations, and across many television channels or video source. For example, it may contain data for all of the television channels distributed in an entire country over the last ten years.
The database can be searched according to some specific rules. For example, it is possible to search the archive and extract the entries for a specific location or for specific time duration.
Before further describing how to make use of the fingerprint archive database, we first discuss the basic processing components with the data center 4. This is shown in
The data center 4 typically operates as follows. A user submits a video content clip 14, which is for specific time duration of video content. The video clip 14 preferably is in digital compressed video data format. The converter 21 takes the video clip 14 as input and performs decompression and passes the resulting digitized video frame data to the fingerprint extractor 22, which obtains the fingerprint samples for the video frames transferred from the converter 21. The output of the fingerprint extractor 22 contains the fingerprint samples associated with the video clip 14.
The fingerprint extractor 22 preferably operates as shown in
Illustration below will focus on the internal operations of the fingerprint extractor 22 in some greater detail, see
In
In the preferred embodiment, each video frames 100 are sampled exactly the same way. In other words, image samples from the same positions are sampled for different images, and the same number of samples is taken from different images. In addition, the images are sampled consecutively.
The samples are then organized as part of the continuous streams of image samples and be placed into the transfer buffer 205. The image samples from different frames are organized together into the transfer buffer 205 before it's sent out. Sampling on images may be performed non-consecutively. In other words, the number of samples taken from each image may be different.
Separately, as shown in
Next it is to show how the selected entries from the archive database 400 can be prepared for the matching operation with the video clip 14. This is illustrated in
Fingerprint archive database 400 holds entries for fingerprint streams from many locations, channels and over potentially very long period of time durations. Fingerprint entries can be retrieved from the database according to location, channel, and time. Fingerprint entries selected meet a specific attributes. Further search among this data results in video information meeting the same attributes.
The fingerprint data segments are then copied out of the entries and assembled into a continuous fingerprint data stream. This stream is the restored output of the sub-sampler within fingerprint extractor 22 shown in
The matcher or correlator 600 in
The matcher 600 takes in two fingerprint data sets. The first is the finite duration fingerprint obtained from the input video clip 14. The second is the fingerprint stream reconstructed from the searched fingerprint archive database 400. The matching result between the two, combined with the additional information obtained from the archive entries, such as time, location, channel and content types, are then put together as the search report. The output of the formatter 204 therefore contains information on when and where the original video clip 14 appeared in the video signals captured by the remote collectors 2.
More specifically, as shown in
Once the fingerprint reconstruction is complete, it is aligned and compared against the fingerprint data obtained from the searching video clip (steps 803, 804). If the result is a match, it means that the two fingerprints represent two visually identical pieces of video content. In this case, the additional information obtained form the data entries, such as location, channel, time and any optional information, will be combined with the information on the searching video clip to product a single report message (step 805). If the two fingerprints do not match, then the fingerprint array obtained from (step 802) is advanced by one frame relative to the searching fingerprint (step 806), and the corresponding information obtained from the newly included fingerprint data points will be updated as well (step 807). The process is then repeated at (step 803).
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/CN2008/071023 | 5/21/2008 | WO | 00 | 5/30/2008 |