Video Super-Resolution Frame Estimation With Ghosting Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In super-resolution of moving images, determining the accuracy of information contained in each frame is challenging when using past frames for estimation, leading to difficulties in achieving high-quality image estimation.
Innovation Solution
An image processing system that includes a processor, memory, and display, utilizing a machine learning model to estimate frames with increased pixel count and quality by incorporating information from past frames, and applying corrections to prevent ghost phenomena and improve image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If information from past frames is used to estimate high-quality moving images, then image quality is improved, but it becomes difficult to determine the accuracy of information in each frame
Solution Approach 1:
The patent introduces an inspection image generated by synthesizing past frame information as an intermediary element. This inspection image serves as a mediator to evaluate the accuracy of input frame information without directly compromising the quality estimation process. The machine learning model processes both the quality estimation and accuracy inspection separately, allowing high-quality output while maintaining verification capability.
2Adaptability or versatility
If attached information is edited according to user instructions, then flexibility and control are improved, but system complexity increases
Solution Approach 1:
The patent implements dynamic editing of attached information where the system adapts to user instructions in real-time. The editor modifies attached information (such as motion vectors or metadata) based on user input, and the machine learning model dynamically adjusts its processing accordingly. This dynamic approach allows flexible user control while managing complexity through modular architecture where only specific components are affected by editing operations.
Data Source
AI summary
Techniques include storing, by a memory, first to Nth (N is a natural number equal to or greater than 2) input frames having a predetermined number of input pixels, and attached information of each of the input frames. The techniques further include editing, by a processor, nth (n=2, 3, . . . , N) attached information stored in the memory according to user instructions. The techniques further include outputting, by a machine learning model trained using a plurality of training data sets, an nth estimated frame having a number of estimated pixels greater than or equal to the number of input pixels, based on first to n−1th input frames and the attached information, and an nth input frame and edited attached information. The techniques further include displaying the nth estimated frame.