Viewpoint Image Generation Using Motion Vectors and Homography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for generating new viewpoint images from multiview input images often result in information loss due to cropping, especially when calibration and warping are performed, which can distort the images and introduce black regions.
Innovation Solution
An image processing method that calculates motion vectors based on homography and disparity information to generate new viewpoint images, using calibration and rectification information to align and transform input images without the need for initial cropping, thereby minimizing information loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If calibration and warping are performed on input images to generate new viewpoint images, then the images can be aligned based on camera environments, but black regions are introduced and information is lost due to subsequent cropping
Solution Approach 1:
The patent performs preliminary actions by calculating motion vectors and generating a new viewpoint image template BEFORE performing cropping operations. The template is created using homography information and disparity information to predict where content should appear, allowing the system to prepare for cropping while minimizing information loss by knowing in advance what regions will be retained.
Solution Approach 2:
The patent introduces a new viewpoint image template as an intermediary between the calibrated/wrapped images and the final cropped output. This template serves as a mediator that guides the cropping process, ensuring that only necessary regions are cropped while preserving important image information. The template acts as a blueprint that prevents arbitrary cropping.
2Object-generated harmful factors
If cropping is performed to remove black regions from calibrated and warped images, then the black regions are eliminated, but image information is lost
Solution Approach 1:
The patent calculates motion vectors and generates the new viewpoint image template before cropping operations. This preliminary calculation identifies which regions contain valid image content versus black regions, allowing the cropping process to remove only necessary black areas while preserving important image information that would otherwise be lost.
Solution Approach 2:
The patent uses the generated new viewpoint image template as feedback to guide the cropping process. The template provides information about where valid content should be located, allowing the system to make informed decisions about which regions to retain and which to crop, thereby minimizing information loss while still removing harmful black regions.
3Manufacturing precision
If motion vectors are calculated based on homography and disparity information, then accurate alignment is achieved, but computational complexity increases
Solution Approach 1:
The patent segments the complex calculation process into distinct components: first calculating homography information from calibration data, then calculating disparity information from rectified images, and finally combining these to generate motion vectors. This segmentation allows each component to be optimized independently and makes the overall complex process more manageable and efficient.
Solution Approach 2:
The patent performs preliminary calculations of homography and disparity information before the main motion vector calculation. By preparing these intermediate results in advance, the system avoids redundant computations during the final image generation phase, reducing overall computational complexity while maintaining high accuracy.
Data Source
AI summary
At least one example embodiment discloses an imaging apparatus that may receive input images with different viewpoints, obtain information associated with the input images, and generate a new viewpoint image based on the received input images and the obtained information and thus, reduce an information loss of the new viewpoint image for the input images.