Video Foreground Completion via Background Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video processing methods using deep learning for foreground completion in video applications face challenges in ensuring real-time performance due to high calculation amounts, making it difficult to achieve real-time foreground completion.
Innovation Solution
A video processing method that acquires a target image in response to a triggering instruction, performs completion filling on the target region using the background region in the target image, and displays the processed image as the current video frame, effectively removing the target object, thereby ensuring both completion effect and real-time performance by processing only two adjacent frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deep learning methods (optical flow network and GAN) are used for foreground completion, then completion quality is improved, but calculation amount increases making real-time performance difficult to ensure
Solution Approach 1:
The patent extracts and removes the target object from the current video frame by identifying it against the background, then fills the resulting region using background information from previous frames. This extraction approach avoids the need for complex deep learning completion while achieving the desired effect of removing the target object in real-time.
Solution Approach 2:
The patent performs preliminary actions by capturing and storing background information from previous video frames before the target object needs to be removed. This pre-captured background data is then used to quickly fill the target region when removal is needed, eliminating the need for computationally intensive real-time completion processing.
2Manufacturing precision
If multiple video frames are processed for foreground completion, then completion effect is improved, but processing time increases
Solution Approach 1:
The patent extracts the target object from the current frame and fills the resulting region using background information from a single previous frame rather than processing multiple frames. This extraction and single-frame filling approach reduces processing time while maintaining completion effect quality.
Solution Approach 2:
The patent performs preliminary capture of background information from previous frames, storing it for later use. When target object removal is needed, this pre-captured background is directly applied to fill the target region, eliminating the need for time-consuming multi-frame processing at the moment of removal.
Data Source
AI summary
Provided in the present disclosure are a video processing method and apparatus, and a device, a medium, a program product and a computer program. The video processing method provided in the present disclosure comprises: in response to a trigger instruction, acquiring a target image, wherein the target image is a video frame prior to a reference image, and the reference image is a video frame that is currently acquired by an image sensor; then performing, by using a background region in the target image, completion and filling on a target region, in which a target object is located, in the reference image, and finally, displaying the processed image as the current video frame.