Scene-Based Video Keyword Generation for Accurate Highlight Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video processing methods suffer from poor correlation between keywords and highlights, and inaccurate descriptions when searching for highlights in videos, leading to inefficient and inaccurate retrieval of desired content.
Innovation Solution
A video processing method that edits videos based on scene changes, extracts feature parameters, and generates keywords associated with the edited target videos to improve correlation and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If keywords are extracted from video content without scene-based segmentation, then the extraction process is simpler and faster, but the correlation between keywords and video highlights deteriorates and description accuracy decreases
Solution Approach 1:
The video is segmented into multiple scenes based on scene change detection algorithms. Each scene is then independently processed to extract feature parameters and generate keywords. This segmentation ensures that keywords are contextually relevant to specific video segments, improving correlation accuracy while managing complexity through modular processing.
Solution Approach 2:
Scene change detection and video segmentation are performed as preliminary steps before keyword extraction. By pre-organizing the video into distinct scenes, the system establishes a structured foundation that facilitates more accurate and efficient subsequent keyword generation, avoiding the need for complex post-processing corrections.
2Measurement precision
If feature parameters are extracted from entire video sequences, then the extraction process is more straightforward, but the identification accuracy of feature parameters deteriorates due to mixed scene content
Solution Approach 1:
The video sequence is divided into multiple independent scene segments based on detected scene changes. Feature parameter extraction is then performed on each scene segment separately rather than on the entire video. This segmentation isolates mixed scene content, allowing for more accurate identification of features within each homogeneous scene while managing complexity through systematic processing of individual segments.
3Measurement precision
If video processing is performed without scene-based editing, then the processing time is shorter, but the description accuracy and retrieval efficiency deteriorate
Solution Approach 1:
The video processing pipeline is segmented into distinct stages: scene change detection, scene segmentation, feature parameter extraction, and keyword generation. This modular segmentation allows each stage to be optimized independently, improving overall description accuracy while managing processing time through efficient, targeted operations on segmented data rather than processing the entire video as a single unit.
Solution Approach 2:
Scene change detection and video segmentation are performed as preliminary actions before the main feature extraction and keyword generation processes. By pre-organizing the video content into scene segments, the system reduces the complexity of subsequent processing steps and enables more accurate description generation, ultimately improving retrieval efficiency despite the additional initial processing time.
Data Source
Figure 1~3
Figure 4~6
Figure 7~8
AI summary
A video processing method, comprising: according to the scenario, editing a video to be edited, and obtaining a target video (S100); acquiring feature parameters of the target video (S200); generating, according to the feature parameters, a keyword of the target video (S300); and associatively storing the keyword and the target video (S400).