Video Logo Detection via Confidence Mask Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video fingerprinting methods are inadequate for identifying videos that do not have a matching entry in their reference database, particularly in high-volume video hosting sites, which can lead to non-compliance with licensing and copyright laws due to incorrect identification of video sources.
Innovation Solution
A computer-implemented method that detects proprietary rights logos in videos by generating a confidence mask using scaling factors for video regions, analyzing generic logo features, and comparing interest point descriptors to stored reference logos, thereby identifying known logos and their positions, scales, and orientations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If video fingerprinting methods are used to identify videos, then videos with matching entries in the reference database can be identified, but videos without matching entries in the reference database cannot be identified
Solution Approach 1:
The system performs preliminary action by extracting and storing feature vectors from logo regions of videos before they are uploaded to the reference database. When a new video is received, the system extracts logo features and compares them against the pre-stored feature vectors to identify the source, enabling identification of videos that would otherwise be unknown in the database.
Solution Approach 2:
The system uses an intermediary approach by introducing logo feature extraction as a mediator between the uploaded video and the reference database. Instead of directly comparing video fingerprints, the system extracts logo features from the video and matches them against stored logo feature vectors, serving as an intermediate representation that enables identification of new videos.
2Reliability
If conventional video fingerprinting is used, then existing videos can be identified, but the system cannot comply with licensing and copyright laws for new videos
Solution Approach 1:
The system extracts the logo regions from the video frames and separates them from the rest of the video content. By taking out only the relevant logo information and storing its feature vectors, the system enables rapid compliance checking of new videos without processing the entire video, thus maintaining high productivity while improving compliance reliability.
Solution Approach 2:
The system performs preliminary extraction and storage of logo feature vectors from reference videos before they are uploaded to the database. This preliminary action enables the system to quickly match new videos against the pre-processed logo features, ensuring compliance with licensing and copyright laws while maintaining efficient processing speeds.
3Adaptability or versatility
If the reference database is expanded to include more videos, then more videos can be identified, but the system becomes less efficient at processing high-volume uploads
Solution Approach 1:
The system extracts only the logo feature vectors from videos and stores them in the reference database, rather than storing complete video data. This extraction approach allows the database to grow in coverage while maintaining efficient processing, as the system only needs to compare logo features rather than entire videos during identification.
Solution Approach 2:
The system performs preliminary extraction of logo features and storage of their vectors before videos are uploaded to the database. This preliminary processing creates a compact reference structure that enables rapid matching of new videos, allowing the database to expand in coverage without proportionally reducing processing efficiency.
Data Source
AI summary
Proprietary rights logos are detected in a video. The video is divided into a plurality of regions that are analyzed for generic proprietary rights logo features. A confidence mask is generated that comprises a plurality of scaling factors, each scaling factor corresponding to a region of the video and indicating a likelihood that the corresponding region of the video includes a proprietary rights logo. The scaling factors of the confidence mask are applied to the video data to generate an altered video. The altered video is analyzed to determine a confidence measure that the video includes a reference proprietary rights logo.