Video Synthesis With Mirrored Object Extensions for Richer Motion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video processing methods for improving video interest are simple and lack richness, resulting in low propagation speed and utilization rate of videos.
Innovation Solution
A video synthesis method that involves obtaining a region image and a background image from a video frame, extending these images in different directions to create extended images, and rendering them to produce a target rendered image, which is then synthesized into a target video.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple video processing methods are used to improve video interest, then the processing complexity is low, but the video propagation speed and utilization rate are low
Solution Approach 1:
The patent segments the video frame into multiple regions including the first region containing the original object, a second region with a mirror image, and additional regions with extended objects. This segmentation enables complex visual effects to be achieved through systematic region division and independent processing of each region, resolving the contradiction between video interest enhancement and processing complexity
Solution Approach 2:
The patent creates mirror images and extended copies of the target object in different regions of the video frame. By generating multiple copies (original object, mirror image, extended objects) and arranging them in specific spatial relationships, the patent enhances video interest through clone effects while maintaining efficient processing through automated copying operations
2Adaptability or versatility
If video processing adds simple content like bullet-screens and stickers, then the processing is easy, but the video interestingness is insufficient
Solution Approach 1:
The patent extends the target object from a single instance to multiple instances arranged in spatial dimensions (horizontal extension, vertical extension) and creates mirror images across different spatial planes. This dimensional expansion transforms a simple object into a complex multi-object scene, significantly enhancing video interestingness while the automated extension process maintains processing efficiency
Solution Approach 2:
The patent creates asymmetric arrangements by generating mirror images (symmetric transformation) and then extending objects in specific directions to create asymmetric compositions. The combination of symmetric mirror imaging and asymmetric extension patterns adds visual complexity and interest to the video content
3Productivity
If the number of objects in motion is increased to enhance video interestingness, then the video synthesis efficiency improves, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing the video frame to identify the target object, generate its mirror image, and determine extension regions before the actual video synthesis. This preliminary preparation of all necessary image elements (original object, mirror image, extended objects) enables efficient video synthesis in the subsequent stage, resolving the contradiction between increased object count and processing complexity
Data Source
AI summary
Embodiments of the disclosure provide a video synthesis method, apparatus, device, medium, and product. The method includes: obtaining a region image corresponding to a target object and a background image excluding the region image by performing an image segmentation on an image frame in a video to be processed; determining a mirror image corresponding to the region image; obtaining at least one first extended image by extending the region image in a first extension direction; obtaining at least one second extended image by extending the mirror image in a second extension direction different from the first extension direction; rendering the background image, the at least one second extended image and the at least one first extended image to obtain a target rendered image at the end of the rendering; obtaining a target video corresponding to the video to be processed by performing a video synthesis on the target rendered image.