Cropped-Region Perspective Correction for Video Stabilization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to effectively correct perspective distortion and stabilize images, leading to blurry and low-quality videos due to terminal shaking and inconsistent image alignment, resulting in a noticeable 'jello' effect.
Innovation Solution
Perform stabilization and perspective distortion correction on cropped areas of images rather than the entire image, ensuring consistent transformation of the same object across frames to maintain inter-frame consistency and reduce the 'jello' effect.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If perspective distortion correction is performed on the entire original image, then the perspective distortion is corrected, but the jello phenomenon occurs due to inconsistent transformation across frames
Solution Approach 1:
The patent divides the image processing into two distinct stages: first performing stabilization to obtain a cropped area, then performing perspective distortion correction only on that cropped area. This segmentation separates the correction process from the entire image, applying transformations only where needed, thereby maintaining consistency across frames while correcting distortion.
Solution Approach 2:
The patent applies perspective distortion correction locally to the cropped area rather than globally to the entire original image. By limiting the correction to the stabilized cropped region, the transformation parameters remain consistent across adjacent frames, preventing the jello phenomenon while still achieving effective distortion correction in the visible area.
2Reliability
If stabilization is performed on the entire image, then the shake is reduced, but the processing complexity and time increase
Solution Approach 1:
The patent extracts only the necessary cropped area from the original image after stabilization, rather than processing the entire image. This extraction reduces the amount of data that needs to undergo subsequent perspective distortion correction, thereby reducing processing complexity and time while maintaining video stability.
Solution Approach 2:
The patent performs stabilization and perspective distortion correction only on the cropped area that is actually needed for the final output, rather than processing the entire original image. This partial action approach reduces computational load and processing time while achieving the same stability effect.
3Stability of the object's composition
If perspective distortion correction is performed after stabilization, then the cropped area maintains consistency, but the processing order adds complexity
Solution Approach 1:
The patent performs stabilization as a preliminary action before perspective distortion correction. By first stabilizing the image to obtain a consistent cropped area, and then applying perspective distortion correction to that stabilized region, the method ensures that the cropped area maintains consistency across frames. This sequential approach, while adding a step, simplifies the overall logic by establishing a clear processing hierarchy.
Data Source
AI summary
This application provides an image processing method, including: performing stabilization on two adjacent image frames of a video stream acquired by a camera to obtain cropped areas, and performing perspective distortion correction on the cropped areas obtained through stabilization. Because the cropped areas output after stabilization is performed on the two adjacent frames include same or basically same content, it can be ensured that a subsequently generated video is smooth and stable. In this application, position relationships, for perspective distortion correction, between the cropped areas obtained through stabilization are further mapped, so that degrees or effects of perspective distortion correction of the adjacent frames of the subsequently generated video are the same, in other words, degrees of transformation of a same object in the images are the same.