Panorama Image Synthesis Using Linearization Coefficients for Uniform Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional panorama image synthesis methods result in non-uniform light source environments, leading to inaccurate object detection, particularly outdoors where light source variations are significant.
Innovation Solution
A method that synthesizes panorama images by calculating a linearization coefficient using pixel values from overlap regions and expanding images to ensure uniform light source environments, allowing for accurate object detection across varying light conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional panorama image synthesis methods are used, then panorama images can be synthesized from multiple images with different shooting directions, but the light source environment becomes non-uniform across the panorama image
Solution Approach 1:
The patent applies parameter changes by calculating a linearization coefficient that transforms images with different light source environments into a unified light source environment. This coefficient adjusts the illumination parameters of each input image to match a reference light source environment, thereby resolving the non-uniformity issue while maintaining the expanded panorama coverage area
Solution Approach 2:
The patent introduces an intermediary element - the linearization coefficient - that acts as a mediator between images with different light source environments and the final panorama image. This coefficient serves as a transformation bridge that harmonizes the illumination characteristics across all input images, enabling uniform light source environment in the synthesized panorama
2Ease of operation
If panorama images with non-uniform light source environments are used for object detection, then object detection can be performed, but detection accuracy is significantly reduced due to false detection of background regions
Solution Approach 1:
By changing the illumination parameters of input images through linearization coefficient calculation, the patent creates panorama images with uniform light source environments. This parameter transformation ensures that background regions maintain consistent appearance across the panorama, enabling accurate object detection without false positives from lighting variations
3Adaptability or versatility
If light source environment varies after panorama image synthesis, then the panorama image can be created, but the difference in light source environment between the panorama image and detection image causes background regions to be detected as objects
Solution Approach 1:
The patent applies preliminary action by performing light source environment unification through linearization coefficient calculation during the panorama synthesis stage, before object detection is performed. This advance normalization ensures that when detection images are later compared against the panorama background, both share compatible light source environments, preventing false detections caused by lighting mismatches
Data Source
AI summary
An overlap image group each having a first overlap region is selected (S11). The overlap image group is composed of N images including a second overlap region other than the first overlap region and a to-be-expanded image not including the second overlap region. A linearization coefficient of the to-be-expanded image is calculated using the images of the overlap image group in the first overlap region (S12), and the to-be-expanded image is expanded in size up to the second overlap region, using the thus calculated linearization coefficient and the N images in the second overlap region (S14). This processing is repeated to generate a panorama image.