Eye-Opening Video Generation Using Gradual Image Transitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video and image processing technologies fail to enhance the interestingness and user experience by not providing dynamic eye opening/closing effects in videos and images.
Innovation Solution
A video generation method that adjusts the eye opening degree of objects in images using a trained image processing model to generate a series of target images with varying eye opening degrees, followed by creating a video with a gradual eye opening or closing effect.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional video recording methods are used, then the video content is simple and straightforward, but the interestingness and user experience are insufficient
Solution Approach 1:
The patent applies preliminary action by pre-training an image processing model with eye opening degree adjustment capabilities before actual video processing. The model is prepared in advance with learned parameters and structures that enable automatic eye opening degree adjustment during video generation, eliminating the need for complex real-time processing during video creation.
Solution Approach 2:
The patent introduces an image processing model as an intermediary between the original image input and the final video output. This model acts as a mediator that automatically adjusts eye opening degrees in intermediate images during video generation, simplifying the overall process while enhancing video interestingness without requiring direct complex manipulation of final video frames.
2Ease of operation
If eye opening degree adjustment is applied to generate dynamic effects, then user experience is improved, but processing time and computational resources increase
Solution Approach 1:
The image processing model is pre-trained in advance with eye opening degree adjustment capabilities. During actual video generation, the model directly applies learned parameters to adjust eye opening degrees without requiring complex real-time computation, significantly reducing processing time while maintaining enhanced user experience.
Solution Approach 2:
The patent changes the parameter approach by using a trained model that directly outputs adjusted eye opening degree values rather than computing them through complex algorithms during video generation. This parameter-based approach reduces computational overhead and generation time while preserving the dynamic eye effects that improve user experience.
3Measurement precision
If multiple training images with different eye opening degrees are used, then the image processing model achieves better accuracy, but the training complexity and data requirements increase
Solution Approach 1:
The patent applies preliminary action by collecting and preparing multiple training images with different eye opening degrees before model training. This pre-prepared dataset enables the model to learn accurate eye opening degree adjustments during training, achieving high precision in the actual video generation process without requiring complex real-time adjustments.
Solution Approach 2:
The patent uses multiple copied variations of training images with different eye opening degrees to train the model. Instead of requiring unique complex datasets, the model learns from replicated training samples with varied eye opening states, reducing data collection complexity while maintaining high adjustment accuracy during video generation.
Data Source
AI summary
The present disclosure relates to a video generation method and apparatus, a device, and a storage medium. After acquiring one or more images to be processed, a plurality of target images having different degrees of eye opening are generated according to the images to be processed, and a video having a process of gradual eye change is generated on the basis of the plurality of target images.