Video signature generation device and method, video signature matching device and method, and program
AssigneeNEC Corporation
InventorsRyoma Oami, Kota Iwamoto
Life of this patent
Estimating the term…
Abstract
A problem of degradation in the accuracy of video matching, which is caused when videos contain video patterns commonly appearing in various videos or video patterns in which features cannot be acquired stably, is solved. In order to solve this problem, a visual feature extraction unit extracts a visual feature to be used for identification of a video based on features of a plurality of pairs of sub-regions in the video, and a confidence value calculation unit calculates a confidence value of the visual feature based on the features of the plurality of pairs of sub-regions. When matching is performed, visual features are compared with each other in consideration of the confidence value.
Claims41
A video signature generation device, for identifying duplicate video content, comprising:
- a processor that controls the video signature generation device to implement:
- a visual feature extraction unit which extracts a visual feature of a video based on features of two sub-regions in a single frame in the video;
- and
- a confidence value calculation unit which calculates a confidence value of the visual feature based on the features of the two of sub-regions in the single frame, wherein the confidence value calculation unit calculates the confidence value based on a difference between the two sub-regions in the single frame.
A video signature matching device that compares (i) a first visual feature of a first video calculated from features of two first sub-regions in a first single frame in the first video to (ii) a second visual feature of a second video calculated from features of two second sub-regions in a second single frame in the second video, and determines a reliability of a result of determining whether the first video is a substantial duplicate of a second video using the features of the two first sub-regions and the two second sub-regions based on the first confidence value information calculated from the features of the two first sub-regions that indicates a first reliability of the first visual feature and second confidence value information calculated from the features of the two second sub-regions that indicates a second reliability of the second visual feature, the device comprising:
- a processor that controls the video signature matching device to implement:
- a matching parameter calculation unit which calculates a matching parameter based on the first confidence value information and the second confidence value information;
- and
- a matching unit which compares the first visual feature with the second visual feature in accordance with the matching parameter, and generates a matching result that indicates whether the first video is the substantial duplicate of the second video.
A matching device comprising:
- a processor that controls the matching device to implement:
- a matching parameter calculating unit that receives a first confidence value calculated from differences between features of two first sub-frames in a first video of a first input signal, the first input signal comprising a first video signature that comprises the first confidence value and a first visual feature of the first video calculated from the features of the two first sub-frames, receives a second confidence value calculated from differences between features of two second sub-frames in a second video of a second input signal, the second input signal comprising a second video signature that comprises the second confidence value and a second visual feature of the second video calculated from the features of the two second sub-frames, and calculates a matching parameter based on the first confidence value and the second confidence value;
- and
- a matching unit that receives the first visual feature, the second visual feature, and the matching parameter, and determines whether the first video is a substantial duplicate of the second video based on the first visual feature, the second visual feature, and the matching parameter.
A video signature generation method for identifying duplicate video content comprising:
- extracting a visual feature of a video based on features of two sub-regions in a single frame in the video;
- and
- calculating a confidence value of the visual feature based on the features of the two sub-regions in the single frame, wherein the calculating comprises calculating the confidence value based on a difference between the two sub-regions in the single frame.
A video signature matching method comprising:
- calculating a matching parameter based on the-first confidence value information and second confidence value information, the first confidence value information calculated from differences between features of two first sub-regions of a first single frame of a first video the first confidence value information indicating a first reliabilitv of a first visual feature of the first video calculated from the features of the two first sub-frames, the second confidence value information calculated from differences between features of two second sub-regions of a second single frame of a second video, the second confidence value information indicating a second reliability of the second visual feature of the second video calculated from the features of the two second sub-frames:
- and
- comparing the first visual feature with the second visual feature in accordance with the matching parameter;
- and
- generating a matching result that indicates whether the first video is the substantial duplicate of the second video based on a result of the comparing.
A matching method comprising:
- receiving a first confidence value calculated from differences between features of two first sub-frames in a first video of a first inpu signal. the first input signal comprising a first video signature that comprises the first confidence value and a first visual feature of the first video calculated from the features of the two first sub-frames;
- receiving a second confidence value calculated from differences between features of two second sub-frames in a second video of a second input signal, the second input signal comprising a second video signature that comprises the second confidence value and a second visual feature of the second video calculated from the features of the hwo second sub-frames;
- calculating a matching parameter based on the first confidence value and the second confidence value;
- and
- obtaining the first visual feature, the second visual feature, and the matching parameter, and determining whether the first video is a substantial duplicate of the second video based on the first visual feature, the second visual feature, and the matching parameter.
A non-transitory computer-readable medium having embodied thereon computer implementable instructions that cause a computer to function as a video signature generation device for identifying duplicate video content, comprising:
- a visual feature extraction unit which extracts a visual feature of a video based on features of two sub-regions in a single frame in the video;
- and
- a confidence value calculation unit which calculates a confidence value of the visual feature based on the features of the two sub-regions in the single frame, wherein the confidence value calculation unit calculates the confidence value based on a difference between the sub-regions in the single frame.
A non-transitory computer-readable medium having embodied thereon computer implementable instructions that cause a computer to function as:
- a matching parameter calculation unit which receives first confidence value information calculated from differences between features of two first sub-frames in a first video of a first input signal, the first input signal comprising a first video signature that comprises the first confidence value and a first visual feature of the first video calculated from the features of the two first sub-frames, and second confidence value information calculated from differences between features of two second sub-frames in a second video of a second input signal, the second input signal comprising a second video signature that comprises the second confidence value and a second visual feature of the second video calculated from the features of the two second sub-frames, and calculates a matching parameter based on the first confidence value and the second confidence value;
- and
- a matching unit receives the first visual feature, the second visual feature, and the matching parameter, and determines whether the first video is a substantial duplicate of the second video based on the first visual feature, the second visual feature. and the matching parameter.
Dependencies are inferred from the claim text.