Camera Overlap Blending for Natural Composite Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for vehicle driving support struggle to create clear, natural composite images from multiple cameras, leading to discontinuity and difficulty in recognizing pedestrians and obstacles near camera overlaps, with previous techniques resulting in unclear images, high hardware demands, and unnatural composites.
Innovation Solution
An image processing system that extracts image feature amounts from multiple camera images and computes blend rates based on correlation, switching blending methods to prioritize the image with stronger feature overlap, allowing for more natural composition and improved visibility of obstacles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If one overhead view image is created by selecting the photographed image to be used for display in accordance with a certain standard, then the processing is simple, but discontinuity arises on the joint and it becomes difficult to recognize pedestrians and obstacles in the vicinity of the joint
Solution Approach 1:
The patent changes the parameter of image selection from fixed standard-based selection to dynamic selection based on correlation calculation. The system calculates correlation between overlapping regions of adjacent camera images and selects the image with higher correlation, thereby maintaining both processing simplicity and recognition accuracy.
2Stability of the object's composition
If pixels are alternately arranged in the region where the plurality of images overlap, then continuity is improved, but the image becomes unclear and the solid object is displayed as a double image
Solution Approach 1:
The patent applies local quality by treating overlapping regions differently from non-overlapping regions. Instead of uniformly alternating pixels across all overlapping areas, the system calculates correlation for each overlapping region and selectively selects images based on correlation strength, thereby maintaining both continuity and clarity in different local areas.
3Reliability
If the joint portion is moved or boundary line is set to avoid obstacles, then recognition of obstacles is improved, but discontinuity arises on the end of photographing range or at the boundary line portion
Solution Approach 1:
The patent introduces dynamics by making the boundary line adaptive rather than fixed. The boundary line automatically adjusts its position based on obstacle detection and correlation calculations, moving away from obstacles while maintaining continuity through correlation-based image selection in the adjusted boundary regions.
4Reliability
If weighted composition is used to prioritize obstacle images, then obstacle visibility is improved, but the system requires high hardware capacity and becomes expensive
Solution Approach 1:
The patent replaces expensive, high-capacity hardware requirements with a computationally efficient software-based correlation calculation approach. By using simple correlation metrics and selective image copying instead of complex weighted composition algorithms, the system achieves obstacle visibility improvement without requiring high hardware capacity.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An objective of the present invention is to generate a more natural composite image wherein a solid object (obstacle) is easily seen. An image detection unit (106) extracts image feature values from images of each camera (101). An image conversion unit (105) computes a blend rate according to the image feature values and composites an image of a superposition area wherein a plurality of camera images overlap. An assessment is made of a correspondence of the image feature values of each image in the superposition area, and a determination is made that a solid object is present if the correlation is weak. Furthermore, a determination is made that the solid object is present in the superposition area if the image feature values in each image have locationally overlapping portions. In such a circumstance, the image is composited with the blend rate of the image with a greater image feature value set large.