Subregion Feature Vector Comparison for Video Modification Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video identification methods cannot accurately determine local differing regions between modified videos, requiring manual playback to identify changes, thus increasing workload.
Innovation Solution
A differing region detection system that generates inter-image difference information by comparing feature amount vectors for subregions between images and detects differing regions based on these differences, outputting detection information to identify local modifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video identity judgment is performed using feature amount vector comparison, then video identity can be determined even with modifications, but local differing regions cannot be identified requiring manual playback
Solution Approach 1:
The patent divides the video frame into multiple subregions and compares feature amount vectors for each subregion between source and modified videos. This segmentation allows identification of specific local differing regions (such as telop positions) while maintaining overall video identity judgment capability.
Solution Approach 2:
The patent assigns different weights or attention levels to different subregions based on their importance for identity judgment. By focusing on subregions with significant feature differences, the system can identify local modifications without being distracted by irrelevant areas, thus solving the local differing region detection problem.
2Measurement precision
If manual playback is used to check differing regions, then local modifications can be identified, but workload increases significantly
Solution Approach 1:
The system automatically performs the comparison of feature amount vectors across subregions and generates difference information without requiring human intervention. The automated detection process eliminates the need for manual playback while maintaining accurate identification of local differing regions, thus reducing workload significantly.
Solution Approach 2:
The patent replaces the mechanical process of manual video playback and visual inspection with an automated computational system that uses feature amount vector comparison and difference information generation to detect differing regions, thereby eliminating time-consuming manual checking.
3Adaptability or versatility
If feature amount vectors are compared for entire images, then overall identity can be judged, but local differences are lost in the comparison
Solution Approach 1:
The patent segments the image into multiple subregions and computes feature amount vectors for each subregion separately. This allows the system to maintain overall identity judgment capability while simultaneously detecting local differences in specific subregions, resolving the contradiction between global and local analysis.
Solution Approach 2:
The patent transitions from comparing entire image feature vectors to comparing subregion feature vectors, adding a spatial dimension to the comparison. This dimensional change enables the system to identify both overall identity and local differing regions by analyzing feature differences at the subregion level.
Data Source
AI summary
The present invention enables detection of a local differing region between images. Inter-image difference information indicating a difference in feature amounts for each subregion between first and second images is generated based on a first feature amount vector that is a set of feature amounts respectively corresponding to a plurality of subregions in the first image and a second feature amount vector that is a set of feature amounts respectively corresponding to a plurality of subregions in the second image, a differing region that is an image region that differs between the first and second images is detected based on differences in the respective subregions indicated by the inter-image difference information, and detection information that indicates a result of the detection is outputted.