Video Piracy Detection via Object Meta-Feature Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic detection systems face difficulties in identifying pirated content, such as hacked videos, due to manipulations like aspect ratio changes, cropping, and format transcoding, which render simple comparison ineffective.
Innovation Solution
A system and method that analyze video frames by identifying recognizable objects, creating vectors based on these objects, and applying statistical methods to determine the correlation between reference and suspect video files, thereby assessing the degree of confidence in identifying pirated content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple comparison methods are used for content detection, then the detection process is fast and simple, but the detection accuracy deteriorates due to manipulations like cropping, rotation, and format transcoding
Solution Approach 1:
The patent segments the video content into individual frames and further into recognizable objects within each frame. By comparing objects frame-by-frame rather than treating the entire video as a single unit, the system can detect manipulations more effectively while maintaining computational efficiency through incremental processing
Solution Approach 2:
The patent introduces meta-features as an intermediary representation layer between the raw video content and the comparison process. These meta-features (color histograms, texture descriptors, shape parameters) serve as robust intermediaries that remain stable under common manipulations like cropping and format transcoding, enabling accurate detection without requiring direct pixel-level comparison
2Measurement precision
If advanced analysis methods are used to handle manipulations, then the detection accuracy improves, but the system complexity increases
Solution Approach 1:
The system divides the complex task of video comparison into manageable segments: frame extraction, object recognition, meta-feature extraction, and vector comparison. This segmentation allows each component to be implemented with relatively simple algorithms while achieving high overall detection accuracy through their coordinated operation
Solution Approach 2:
The patent transforms video content into a different parameter space by extracting meta-features that describe objects in terms of color distributions, texture patterns, and geometric properties rather than raw pixel values. This parameter transformation simplifies the comparison task while making the detection robust to common manipulations
3Object-affected harmful factors
If manipulations like cropping and format transcoding are applied to pirated content, then the pirated content becomes harder to detect, but the recognizable objects within the content remain intact
Solution Approach 1:
The patent extracts recognizable objects from their original video context and represents them through meta-features that capture their essential characteristics independently of the surrounding content. By taking out the objects and comparing their extracted features rather than the full video context, the system can identify pirated content even when portions have been cropped or format-transcoded
Solution Approach 2:
The patent uses color histograms as meta-features to represent objects, which capture the distribution of colors within objects rather than specific pixel values. This approach makes the detection robust to color space transformations and format transcoding while maintaining the ability to identify the same visual content
Data Source
AI summary
Methods and systems for comparing content are provided. A recognizable object which appears in frames of a reference video file is recognized. A meta-feature of the recognizable object is determined. A first vector indicative of the meta-feature of the identified recognizable object in the reference video file is created. The meta-feature of the identified recognizable object which appears in a suspect video file is determined. A second vector indicative of the meta-feature of the identified recognizable object in the suspect video file is created. A correlation between the first vector and the second vector is determined. The correlation between the first vector and the second vector is indicative of a degree of confidence that the suspect video file is a copy of the reference video file.