Image Stitching Using Reused Projection Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image stitching methods for CCTV systems are time-consuming and cannot meet the requirements for real-time video stream processing due to high calculation complexity, particularly in the image registration phase.
Innovation Solution
The method reuses projection transformation parameters and exposure compensation parameters from previous frames when the relative position and brightness conditions between cameras remain unchanged, allowing for parallel calculation of new parameters if conditions change, thereby reducing calculation complexity and processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image stitching methods are used with complete image registration and composition procedures, then stitching accuracy is maintained, but processing time becomes too long for real-time video stream processing
Solution Approach 1:
The patent performs image registration and calculates projection transformation parameters only for the first frame in advance. For subsequent frames, it reuses these pre-calculated parameters without re-registration, thereby eliminating redundant calculations while maintaining stitching accuracy. This preliminary action resolves the contradiction by preparing necessary transformation data before real-time processing requirements arise.
Solution Approach 2:
The patent applies partial action by performing complete image registration only when necessary (first frame or when camera positions change), and using simplified parameter reuse for subsequent frames. This selective approach maintains stitching accuracy where needed while reducing processing time during normal operation, achieving real-time performance without sacrificing quality.
2Measurement precision
If image registration is performed for every frame to ensure accurate stitching, then stitching quality is maintained, but calculation complexity increases significantly
Solution Approach 1:
The patent performs the complex image registration and homography matrix calculation as a preliminary action for the first frame only. These calculated parameters are then stored and reused for subsequent frames, avoiding repeated complex calculations. This approach maintains stitching quality through accurate initial registration while dramatically reducing calculation complexity for real-time processing.
Solution Approach 2:
The patent creates a copy of the projection transformation parameters (homography matrix) calculated from the first frame and reuses this copied data for subsequent frames. Instead of recalculating complex registration parameters for every frame, the system copies and applies the proven parameters, maintaining stitching quality while reducing computational burden.
3Measurement precision
If projection transformation parameters are recalculated for each frame to adapt to position changes, then stitching accuracy is maintained, but processing speed decreases
Solution Approach 1:
The patent implements a dynamic parameter update strategy where projection transformation parameters are recalculated only when camera positions actually change (detected through feature point matching), rather than recalculating for every frame. This dynamic approach maintains stitching accuracy when changes occur while achieving high processing speed during stable conditions, resolving the speed-accuracy contradiction.
Solution Approach 2:
The patent employs periodic action by performing complete image registration and parameter recalculation at specific intervals (when camera position changes are detected), rather than continuously for every frame. Between these periodic updates, the system reuses existing parameters, maintaining stitching accuracy when needed while achieving sustained high processing speed during normal operation.
Data Source
AI summary
Various embodiments of the teachings herein include an image stitching method comprising: acquiring a current frame of a first picture photographed by a first camera and a second camera, wherein an overlap exists between photographed regions in the pictures; determining whether the relative position relationship between the cameras stay unchanged when a current frame and a previous frame of pictures are photographed; if the first condition is met, acquiring a first projection transformation parameter used when a projection transformation is performed for the previous frame of a third picture photographed by the first camera and a second projection transformation parameter used when a projection transformation is performed for the previous frame of a fourth picture photographed by the second camera; and blending the pictures using the first parameter to perform a projection transformation for the first picture and the second parameter to perform a projection transformation for the second.