Foreground Information Generation for Neural Network Video Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face challenges in maintaining recognition accuracy while reducing processing speed, particularly due to camera movements and noise extraction issues when processing video frames.
Innovation Solution
An image processing apparatus and method that generates foreground information indicating the target region within a video frame and applies this information, along with frame differences, to a neural network model trained on image feature maps, allowing for efficient processing by distinguishing key and non-key frames and managing noise effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all video frames are input to a machine neural network model for object extraction, then recognition accuracy is improved, but processing speed decreases and processing load increases
Solution Approach 1:
The patent segments video frames into key frames and non-key frames based on motion detection. Key frames (where significant motion is detected) are processed by the full neural network model for high accuracy, while non-key frames use simplified processing. This segmentation allows the system to maintain high recognition accuracy for important frames while improving overall processing speed by reducing computation on less critical frames.
2Productivity
If difference information is extracted for non-key frames to reduce processing load, then processing speed is improved, but recognition accuracy decreases due to camera movement and noise
Solution Approach 1:
The patent introduces a motion detection module as an intermediary that identifies key frames requiring full neural network processing. This mediator determines which frames need accurate object extraction (key frames with significant motion) versus which can use simplified processing (non-key frames), thereby maintaining recognition accuracy for important frames while improving overall processing efficiency.
Solution Approach 2:
The patent dynamically changes processing parameters based on frame characteristics. For key frames, the full neural network model is applied with high computational resources. For non-key frames, simplified processing is used. The threshold for determining key frames can be adjusted to balance between processing speed and recognition accuracy, allowing flexible adaptation to different scenarios.
Data Source
AI summary
An image processing apparatus includes: a foreground information generating unit that generates, from a frame constituting video frame, foreground information indicating a region of the frame in which a target is present; and a model applying unit that applies the frame and the foreground information generated from the frame to a neural network model that has performed machine learning of an image feature map of the target.