Signal Processing Circuit for Multi-Camera Image Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal processing methods for stitching images from multiple camera modules often result in errors in overlapping regions, especially when the overlapping area is reduced, leading to unstable and disagreeable viewing experiences, particularly with moving images.
Innovation Solution
A signal processing circuit that segments images into overlapping, direct-transformation, and buffer regions, using camera initial conditions and composition conditions to stitch these regions separately, with a composition-condition estimation unit to refine the stitching process, especially utilizing high-resolution and high-frame-rate data from overlapping regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If the overlapping region is made smaller to widen the photographic angle-of-view, then the photographic angle-of-view is improved, but errors between feature points remarkably appear on the opposite side of the overlapping region
Solution Approach 1:
The patent divides the image into three distinct regions: overlapping region, buffer region, and direct-transformation region. This segmentation allows different processing strategies to be applied to each region, enabling accurate stitching even when the overlapping region is small. The buffer region specifically addresses the problem of error propagation by providing a transition zone between the overlapping and direct-transformation regions.
Solution Approach 2:
The patent applies different transformation methods to different regions: feature point-based composition for the overlapping region, intermediate composition conditions for the buffer region, and direct camera initial conditions for the direct-transformation region. This local differentiation ensures that each region is processed with the most appropriate method, maintaining accuracy throughout the entire image even when the overlapping region is minimized.
2Productivity
If feature point extraction is used for stitching, then stitching processing can be performed, but errors between feature points cause the angle-of-view to change slightly and quickly in moving images
Solution Approach 1:
The patent performs preliminary extraction of camera initial conditions (directions and magnifications) from the camera modules before stitching. These pre-extracted parameters are then used as reliable reference data for transforming the direct-transformation region, avoiding the instability that would result from relying solely on feature point extraction during runtime processing.
Solution Approach 2:
The patent introduces buffer regions with intermediate composition conditions that act as mediators between the overlapping region (processed by feature point matching) and the direct-transformation region (processed by camera initial conditions). This intermediary approach smooths transitions and prevents abrupt changes in angle-of-view, maintaining stability in moving images while preserving processing efficiency.
Data Source
AI summary
The present disclosure relates to a signal processing circuit and an imaging apparatus by which a plurality of images including an overlapping region can be accurately stitched.Stitching (projective transformation and warping) is performed on direct-transformation region images on the basis of camera initial condition information (information on directions of camera modules) stored in an EEPROM. Stitching is performed on overlapping-region images under the determined composition condition. Stitching is performed on buffer region images by the use of a value (e.g., moderate value) determined on the basis of a projective transformation condition of the direct-transformation region images and a projective transformation condition of the overlapping-region images. The present disclosure is applicable to, for example, a signal processing circuit used in an imaging apparatus such as a camera.