High-Resolution Image Generation with Dynamic Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image composition techniques that increase image resolution by combining frames with positional differences often result in unnatural impressions due to changes in position and brightness, limiting the subjects and scenes that can be effectively composite.
Innovation Solution
An image processing apparatus and method that obtain multiple frames of images taken from different viewpoints, generate a high-resolution image by combining these frames, detect changed areas, and produce a final high-resolution image by integrating the combined image with areas from a single frame image corresponding to the changed areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple frames of still images with positional differences are combined to increase resolution, then the resolution of the composite image is improved, but unnatural impressions occur in areas that exhibit changes in position and brightness between frames
Solution Approach 1:
The patent segments the composite image into multiple regions: a first region containing pixels from all input frames (including changed areas with positional differences), and a second region containing pixels from a single reference frame (unchanged areas). This segmentation allows different processing strategies to be applied to different regions, resolving the contradiction by isolating problematic changed areas from the overall composite image while preserving high resolution in stable regions.
Solution Approach 2:
The patent applies local quality by treating different regions of the image differently. The first region (changed areas) is processed to handle positional variations, while the second region (unchanged areas) is processed to maintain natural appearance. This local differentiation allows the system to optimize for both resolution enhancement and naturalness in different parts of the image simultaneously.
2Measurement precision
If multiple frames of still images with positional differences are combined to increase resolution, then the resolution of the composite image is improved, but the subjects and scenes for which intended composite images can be obtained are significantly restricted
Solution Approach 1:
The patent introduces dynamic region classification that adapts to the specific characteristics of each input frame sequence. By dynamically identifying changed areas versus unchanged areas, the system can process diverse subjects and scenes (including those with motion, varying brightness, or positional shifts) while maintaining image quality. This dynamic approach removes the rigid constraints of traditional composite image processing.
Data Source
AI summary
An image processing apparatus obtains a plurality of frames of first images, wherein the first images differ from each other in viewpoint. The apparatus generates one frame of second image having a higher resolution than the first images by combining the plurality of frames of first images, and generates a third image having a same resolution as the second image from one of the plurality of frames of first images. The apparatus detects a changed area that has changed among the plurality of frames of first images and further generates a fourth image, as a high-resolution image based on the plurality of frames of first images, from the second image and an area of the third image corresponding to an area of the second image corresponding to the changed area.