Video Signature Generation Using Selective Frame Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video fingerprinting solutions lack robustness, reliability, and efficiency in identifying and comparing videos, especially in cases of degradation, tampering, or partial video availability.
Innovation Solution
A system generates video signatures using a subset of pixels or color schemes from a sequence of video frames, employing a feature selection module, sequence selection module, and signature generation module, which allows for efficient matching and comparison even in altered or tampered videos by deriving curves from image features and applying hash functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video fingerprinting uses all video frames and features, then identification accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The video is divided into multiple frames, and only selected frames are used for fingerprint extraction. The system segments the video content into discrete frame units and applies feature extraction only to representative frames rather than all frames, reducing computational load while maintaining identification accuracy.
Solution Approach 2:
The system uses a subset of video frames and a subset of image features (such as SIFT features) to generate the video fingerprint, rather than processing all frames and all possible features. This partial action approach achieves sufficient identification accuracy with reduced computational complexity.
2Reliability
If video fingerprinting processes all video frames, then reliability is improved, but processing speed decreases
Solution Approach 1:
The system performs preliminary selection of representative video frames before fingerprint extraction. By pre-identifying key frames that best represent the video content, the system ensures reliable fingerprint generation while avoiding the need to process all frames, thus maintaining processing speed.
Solution Approach 2:
The video stream is segmented into discrete frames, and the system selectively processes only certain frames based on their representativeness. This segmentation allows the system to maintain reliability by capturing essential video characteristics while improving processing speed through selective frame sampling.
3Reliability
If video fingerprinting uses comprehensive image features, then robustness to degradation is improved, but computational efficiency decreases
Solution Approach 1:
The system extracts only the necessary and most discriminative image features (such as SIFT features) from selected video frames, rather than using all possible image features. This extraction of essential features maintains robustness to degradation while improving computational efficiency by avoiding processing of redundant features.
Solution Approach 2:
The system changes the parameter of feature selection by using a subset of distinctive image features rather than comprehensive features. This parameter change allows the system to achieve adequate robustness with lower computational cost by focusing on the most informative features.
4Productivity
If video fingerprinting uses a subset of pixels or color schemes, then computational efficiency is improved, but measurement precision may decrease
Solution Approach 1:
The system extracts specific, meaningful features (such as SIFT features) from selected regions or frames of the video, rather than using all pixels or all color information. This selective extraction of discriminative features maintains signature accuracy while significantly improving computational efficiency.
Solution Approach 2:
The system applies feature extraction to specific local regions or selected frames rather than uniformly processing the entire video. By focusing computational resources on locally significant regions that best represent the video content, the system maintains accuracy while improving efficiency.
Data Source
AI summary
Techniques are disclosed for identifying a video using a video signature generated using image features derived from a portion of the video. In some examples, a method may include determining image features derived from a portion of a video, determining a video frame sequence of the video, and generating the video signature of the video based on the image features and the video frame sequence. The method may further include deriving a curve for the video based on the image features and the video frame sequence, and comparing the derived curve with one or more curves corresponding to respective one or more reference videos.