Video Discontinuity Detection Using Gradient Maps and Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera systems face challenges in distinguishing between motion caused by changes in lighting conditions, compression artifacts, and actual motion within a scene, leading to false positive notifications and increased network bandwidth consumption.
Innovation Solution
The system employs gradient map generation, blockwise sum of squared differences (SSD) calculation, and convolution with an inverted Laplacian of Gaussian kernel to differentiate between rapid and gradual changes in video frames, determining if discontinuities are due to motion, illumination changes, or artifacts, thereby adjusting sensitivity settings and skipping non-motion frames for further processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system sends alerts for all detected discontinuities in video frames, then motion detection sensitivity is improved, but false positive alerts increase due to lighting changes and compression artifacts
Solution Approach 1:
The patent segments the video processing pipeline into distinct analysis stages: gradient map generation, blockwise SSD calculation, convolution filtering, and classification. By dividing the discontinuity detection process into these separate functional blocks, the system can analyze different aspects of frame changes independently and combine results to distinguish true motion from false positives caused by lighting changes or compression artifacts.
Solution Approach 2:
The patent introduces an intermediary classification mechanism that acts as a mediator between raw discontinuity detection and final alert generation. This intermediary layer analyzes the nature of detected discontinuities using gradient maps and convolution operations to determine whether they represent actual motion or artifacts, thereby filtering false positives before alerts are sent to users.
2Measurement precision
If the system processes all video frames for motion detection, then detection thoroughness is improved, but network bandwidth consumption increases due to transmitting unnecessary alert data
Solution Approach 1:
The patent extracts and processes only the essential features needed for motion detection—specifically gradient maps and blockwise SSD values—rather than transmitting or processing entire video frames. By extracting only the relevant discontinuity information and filtering out false positives through convolution analysis, the system reduces the volume of data that needs to be transmitted over the network while maintaining thorough detection capabilities.
3Productivity
If the system uses simple frame comparison for motion detection, then processing speed is improved, but measurement accuracy deteriorates due to inability to distinguish motion from lighting changes
Solution Approach 1:
The patent implements a dynamic, multi-stage processing pipeline that adapts the level of analysis based on the characteristics of detected discontinuities. The system begins with rapid gradient map generation and blockwise SSD calculation for quick initial assessment, then applies convolution filtering and classification only when necessary to resolve ambiguous cases. This dynamic approach maintains high processing speed for clear cases while improving accuracy for ambiguous situations through selective application of more computationally intensive analysis.
Data Source
AI summary
Techniques are generally described for discontinuity detection. A first and second frame of video data representing a portion of a physical environment are received. The first and second frames are divided into first and second blocks of pixels, respectively. A set of matching blocks are determined by matching each block of the first blocks with a corresponding block of the second blocks. A difference value is determined for each pair of matching blocks of the set of matching blocks. A mean difference value for the set of matching blocks. The mean difference value is convolved with a convolution kernel. A determination is made that a result of the convolving the mean difference value with the convolution kernel exceeds a mean threshold value. A determination is made that a discontinuity between the first frame and the second frame is not due to motion in the physical environment.