Multi-band Seam Blending for Panoramic Image Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The combination of images from multiple cameras to create panoramic views often results in image defects and artifacts at the seam between images due to differences in camera orientations and other technical challenges, such as parallax effects and varying focal lengths.
Innovation Solution
The method involves using Laplacian of Gaussian masks and linear blending masks for selected bands in overlapping image portions, with a specific bandwidth calculation based on a sharpness factor, and applying these masks to create blended images that are then upscaled and summed to reduce artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If multiple images are combined to create panoramic views, then the field of view is widened, but image defects and artifacts appear at the seam between images
Solution Approach 1:
The patent divides the overlapping image region into multiple frequency bands (low, mid, high frequencies) and processes each band separately with different blending masks. This segmentation allows targeted handling of different image features - low frequencies for smooth transitions and high frequencies for sharp details - thereby reducing visible artifacts while maintaining the widened field of view benefit
Solution Approach 2:
The patent applies different blending mask types to different frequency bands within the overlapping region. Low-frequency bands use one type of blending mask for smooth transitions, while high-frequency bands use another type to preserve sharp edges and details. This local quality differentiation minimizes artifacts at the seam while maintaining overall image quality across the panoramic view
2Ease of manufacture
If conventional blending methods are used to combine images, then the processing is simple, but visible artifacts and defects remain at the seam
Solution Approach 1:
The patent transforms the blending problem from a single-dimensional approach to a multi-dimensional frequency domain approach. By decomposing the image into multiple frequency bands and applying different processing strategies to each band, the method achieves superior artifact reduction while maintaining computational feasibility through systematic multi-band processing
3Area of stationary object
If different camera orientations and focal lengths are used to capture panoramic images, then the coverage area increases, but incongruities and artifacts appear at the seam
Solution Approach 1:
The patent changes the processing parameters differently for different frequency bands to compensate for variations caused by different camera orientations and focal lengths. By adjusting blending mask characteristics and bandwidth parameters across frequency bands, the method achieves consistent image congruity at the seam while maintaining the expanded coverage area benefit from multiple camera configurations
Data Source
AI summary
A method, apparatus and computer program product are provided to combine images captured by multiple cameras into a panoramic image, such as a 360° panorama, in a manner that reduces image defects and other artifacts at or near a seam between images. In some implementations of example embodiments, the overlapping portions of two images are decomposed into a predetermined number of bands. For most of the bands, a laplacian of Gaussian blending mask is applied when blending images associated with the relevant bands. For a final band, a linear blending mask is applied, and the bands are combined into an image that may be used in connection with a panoramic image featuring portions of multiple images captured by an array of cameras. In some example implementations, a bandwidth for the final band may be determined based in part on a selected sharpness factor.