Image Block Processing for Real-Time Panorama Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating panoramas suffer from low efficiency, particularly in achieving accurate and seamless wide-angle views due to image distortion, seam issues, and the inability to handle real-time high-resolution video streams effectively.
Innovation Solution
A method involving image processing that retrieves and processes image blocks from a storage device using block position information, utilizing burst transmission protocols and parallel processing to generate target images, which are then combined to form panoramas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CPU and/or GPU are used to stitch image sequences according to stitching algorithm, then panorama with wide-angle view can be formed, but the processing efficiency is low and cannot meet real-time high-resolution video stream requirements
Solution Approach 1:
The patent divides the source image into multiple image blocks and processes them in parallel. Each processing core handles specific image blocks independently, transforming coordinate systems and projecting images onto a two-dimensional plane simultaneously. This segmentation approach enables real-time processing of high-resolution video streams by avoiding sequential processing bottlenecks.
Solution Approach 2:
The patent transforms three-dimensional spherical coordinate systems into two-dimensional planar coordinate systems through mathematical projection. This dimensionality change allows efficient processing and stitching of wide-angle images while maintaining image quality and reducing computational complexity compared to pure 3D processing approaches.
2Area of stationary object
If fisheye lens system is used for hyper hemispherical imaging, then wide viewing angle of more than 180° can be achieved, but large distortion occurs and resolution at edges is greatly lower than at center
Solution Approach 1:
The patent applies distortion correction algorithms that modify image parameters to compensate for fisheye lens distortion. By calculating and applying correction factors based on the radial distance from the image center, the system restores uniform resolution across the entire field of view while maintaining the wide viewing angle capability.
Solution Approach 2:
The patent performs distortion correction and coordinate transformation as preliminary processing steps before the actual stitching operation. By pre-processing the images to correct distortion and transform coordinates, the subsequent stitching process becomes simpler and more efficient, avoiding the need for complex real-time corrections during assembly.
3Area of stationary object
If multiple cameras are used for capturing image sequences, then wide-angle panorama can be obtained, but image distortion and angle errors cause obvious seams and overlapping portions
Solution Approach 1:
The patent implements an image registration process that uses feedback from overlapping regions to adjust and optimize the alignment of adjacent images. By analyzing feature points and correspondence in overlapping areas, the system calculates transformation parameters that minimize seams and improve stitching accuracy, ensuring seamless panoramas even with multiple cameras having angle errors.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for image processing are provided in the present disclosure. The method may include obtaining a plurality of source images generated by a plurality of imaging sensors. The method may also include processing each of the plurality of source images by: retrieving a plurality of source image blocks from the source image according to block position information associated with the corresponding imaging sensor; generating, for each of the plurality of source image blocks, a target image block based on the source image block; and forming a target image based on the generated target image blocks. The method may further include generating a combined image based on the target images.