Moving Image Scene Change Detection With Main Frame Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for detecting scene changes in moving images are inefficient in accurately identifying scene change points and key frames, particularly in complex environments like air conditioner installations or repairs, leading to suboptimal learning experiences for unskilled workers.
Innovation Solution
An image processing method that utilizes a server apparatus to detect scene change points and key frames by calculating Bhattacharyya distance and pHash distances between frames, extracting edges, and determining key frames through ORB feature matching, ensuring accurate identification of scene changes and key frames in moving images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple pixel value comparison is used for scene change detection, then the processing is fast and simple, but the detection accuracy is low especially in complex environments
Solution Approach 1:
The patent segments the scene change detection process into multiple stages: candidate detection using simple pixel comparison, candidate filtering using Bhattacharyya distance for color histogram comparison, and final verification using ORB feature matching. This segmentation allows each stage to operate at appropriate complexity levels, maintaining speed in early stages while improving accuracy in later stages.
Solution Approach 2:
The patent introduces intermediate processing steps between simple pixel comparison and final scene change determination. Specifically, it uses Bhattacharyya distance as an intermediary metric to compare color histograms of candidate frames, and ORB feature points as another intermediary layer for geometric feature matching. These intermediaries bridge the gap between simple and complex methods, improving accuracy without completely sacrificing processing speed.
2Measurement precision
If multiple detection methods (Bhattacharyya distance, pHash, ORB matching) are used to improve accuracy, then the detection precision improves, but the device complexity and processing time increase
Solution Approach 1:
The patent divides the detection system into modular components: a candidate detection module using pixel comparison, a color analysis module using Bhattacharyya distance, a hash-based module using pHash, and a feature matching module using ORB. Each module operates independently and can be selectively applied, reducing overall system complexity while maintaining high detection accuracy through the combination of specialized functions.
Solution Approach 2:
The patent applies multiple detection methods selectively rather than uniformly to all frames. It first identifies candidate frames using simple pixel comparison, then applies more complex Bhattacharyya distance and ORB matching only to these candidates. This partial application of complex methods reduces computational burden and system complexity while still achieving high accuracy where needed.
3Measurement precision
If multiple detection methods are applied to all frames, then the detection accuracy improves, but the processing time and energy consumption increase significantly
Solution Approach 1:
The patent segments the frame processing into two groups: all frames undergo simple pixel value comparison for candidate identification, while only candidate frames undergo complex Bhattacharyya distance and ORB feature matching. This segmentation ensures that high-accuracy methods are applied minimally, reducing processing time and energy consumption while maintaining detection accuracy for critical frames.
Solution Approach 2:
The patent applies computationally intensive detection methods (Bhattacharyya distance, ORB matching) partially only to candidate frames identified by preliminary simple comparison, rather than to all frames. This partial application dramatically reduces processing time and energy consumption while maintaining high detection accuracy for the frames that matter most.
Data Source
AI summary
An image processing method, by an image processing apparatus, detects candidates of a scene change point in a moving image, detects candidate frames of a main frame of a scene from frames forming the moving image, and determines the main frame from the candidate frames. Based on a time-series order of the candidates of the scene change point and the main frame, when the main frame does not exist between two candidates of the scene change point that are consecutive in time series, the scene change point with one of the two candidates of the scene change point excluded, and when a candidate of the scene change point exists between two main frames that are consecutive in time series and are similar to each other, the scene change point with the candidate of the scene change point excluded.