Video Frame Filtering for Computational Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video content analysis methods face challenges with high computational complexity, making it difficult to implement hardware computing for all frames of a video, especially when computing power is insufficient.
Innovation Solution
A computing device comprising a preprocessing module, a first deep learning module, and a second deep learning module, which processes frames to determine relevance and delete unnecessary frames, reducing computational complexity by using modules with varying processing speeds and complexities to handle video content analysis efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video content analysis is performed on all frames using existing methods, then analysis accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the video processing task into multiple stages: a lightweight preprocessing module that performs initial frame filtering, followed by deeper analysis modules that process only the filtered frames. This segmentation allows accurate analysis to be applied selectively rather than uniformly to all frames, reducing overall computational complexity while maintaining analysis accuracy for relevant frames.
Solution Approach 2:
The patent applies partial action by performing comprehensive analysis only on a subset of frames that pass the preprocessing filter, rather than analyzing every frame. The preprocessing module performs sufficient filtering to identify relevant frames, and the deeper analysis is applied partially only to those selected frames, achieving accurate analysis where needed without the excessive computational cost of analyzing all frames.
2Measurement precision
If all frames are processed by deep learning modules, then detection accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent implements preliminary action through the preprocessing module that performs initial frame filtering before frames reach the deep learning modules. By pre-identifying and filtering relevant frames in advance, the system prepares the data so that subsequent deep learning processing can focus efficiently on only the frames that require detailed analysis, improving overall processing speed while maintaining detection accuracy.
Solution Approach 2:
The processing pipeline is segmented into a fast preprocessing stage and a slower but more accurate deep learning stage. The preprocessing module quickly filters frames based on basic criteria, separating frames that need detailed analysis from those that don't. This segmentation allows the system to maintain high processing speed overall while ensuring detection accuracy is achieved for the relevant subset of frames.
3Productivity
If hardware computing power is increased to handle all frames, then analysis performance is improved, but device complexity and cost increase
Solution Approach 1:
The patent applies partial action by using hardware computing resources to process only a subset of frames that pass the preprocessing filter, rather than distributing hardware resources to process all frames. This approach achieves high analysis performance on relevant frames using available hardware, without requiring additional hardware complexity that would be needed to maintain the same performance level across all frames.
Solution Approach 2:
The patent changes the parameter of frame selection by introducing a preprocessing filter that modifies which frames are passed to the hardware computing modules. By changing the input parameter (from all frames to filtered frames), the system achieves high analysis performance with existing hardware resources, avoiding the need to increase hardware complexity or power to process the full frame set.
Data Source
AI summary
A computing device for handling video content analysis, comprises a preprocessing module, for receiving a first plurality of frames and for determining whether to delete at least one of the first plurality of frames according to an event detection, to generate a second plurality of frames according to the determination for the first plurality of frames; a first deep learning module, for receiving the second plurality of frames and for determining whether to delete at least one of the second plurality of frames according to a plurality of features of the second plurality of frames, to generate a third plurality of frames according to the determination for the second plurality of frames; and a second deep learning module, for receiving the third plurality of frames, to generate a plurality of prediction outputs of the third plurality of frames.