Video Synchronization via Log-Polar Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital watermarking systems face challenges in synchronizing video files due to geometrical transformations, such as rotation, translation, and scaling, which can lead to inaccurate detection of watermarks, high resource consumption, and vulnerability to attacks.
Innovation Solution
A method and system for synchronizing video files by detecting shot boundaries, extracting features in the log-polar domain, and comparing them using normalized cross correlation to determine a matching index, which helps in accurately synchronizing videos even under geometrical transformations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional digital watermarking systems are used for video synchronization, then the system can detect watermarks in video files, but the accuracy deteriorates due to geometrical transformations such as rotation, translation, and scaling
Solution Approach 1:
The patent transforms video features from the conventional Cartesian coordinate system to the log-polar domain, changing the parameter space representation. This transformation makes the feature extraction invariant to geometrical transformations like rotation, translation, and scaling, thereby maintaining watermark detection accuracy despite these harmful factors
Solution Approach 2:
The patent introduces a new dimensional representation by converting spatial coordinates to log-polar coordinates. This dimensional transformation creates a feature space where geometrical transformations appear as simple shifts rather than complex distortions, enabling accurate synchronization even when videos undergo geometrical modifications
2Productivity
If conventional video synchronization methods are used, then the system can process video files, but resource consumption increases due to the complexity of handling geometrical transformations
Solution Approach 1:
By changing the feature extraction approach to use log-polar domain representation, the patent reduces the computational complexity required to handle geometrical transformations. This parameter transformation allows for more efficient processing as the system no longer needs to perform complex rotation and scaling compensation algorithms
Solution Approach 2:
The patent segments the video processing task into distinct modules: shot boundary detection, feature extraction in log-polar domain, feature matching, and synchronization. This segmentation allows each module to be optimized independently, improving overall processing efficiency while reducing resource consumption
3Reliability
If conventional watermarking systems are used, then the system can embed watermarks in video files, but the system becomes vulnerable to attacks due to lack of robustness against geometrical transformations
Solution Approach 1:
The patent applies preliminary anti-action by pre-transforming video features into the log-polar domain before watermark embedding and detection. This preliminary transformation creates an invariant representation that resists subsequent geometrical transformations and attacks, making the watermarking system robust against manipulation
Solution Approach 2:
The patent provides beforehand cushioning by establishing a robust feature matching mechanism in the log-polar domain that can compensate for geometrical transformations and attacks. This cushioning mechanism ensures that even when videos undergo transformations or attacks, the watermark detection remains accurate and reliable
Data Source
AI summary
The present subject matter discloses systems and methods for synchronizing a plurality of video files. In one implementation, the method comprises detecting a one shot boundary in each of a first video file and a second video file, determining a first reference point in the first video and a second reference point in second video based on the shot boundary detection and extracting a plurality of features associated with the reference points. The method further comprises comparing each of the plurality of corresponding features extracted from the reference points in a log-polar domain based in part on normalized cross correlation, computing a matching index, indicative of the extent of match between the extracted features, and generating a notification indicating the synchronization of the first video file and the second video file if the matching index is determined to be greater than a pre-defined matching threshold.