Stereoscopic Fisheye Image Conversion With Selective Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in distributing stereoscopic images for live streaming in real-time due to the increased processing time required for providing images for the right and left eyes, which is not feasible with conventional frame rates.
Innovation Solution
A dual-lens unit captures images with a predetermined parallax, converting them into equidistant cylindrical format using template matching and interpolation techniques, optimizing processing time and image quality for stereoscopic viewing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If images for right eye and left eye are provided in real-time live streaming, then stereoscopic viewing quality is improved, but processing time increases making real-time distribution difficult
Solution Approach 1:
The image processing is segmented into two distinct processing paths: first image processing that converts fisheye images to equidistant cylindrical images, and second image processing that performs perspective transformation. This segmentation allows parallel processing of different image regions and reduces overall processing time while maintaining stereoscopic viewing quality
Solution Approach 2:
The first image processing (fisheye to equidistant cylindrical conversion) is performed in advance as a preliminary step before the second image processing (perspective transformation). This preliminary action prepares the image data in a format that facilitates faster subsequent processing and enables real-time distribution
2Adaptability or versatility
If two images are captured using two image capturing lenses, then stereoscopic viewing capability is improved, but processing complexity increases
Solution Approach 1:
The processing of right eye and left eye images is merged into a unified processing framework where both images undergo the same two-stage processing pipeline (fisheye to equidistant cylindrical conversion, then perspective transformation). This merging reduces processing complexity by reusing the same processing algorithms and structures for both images rather than developing separate processing paths
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
An image processing apparatus acquires an image in which first and second image regions, each of which is a region of a fisheye image, are arranged, and converts the first and second image regions into third and fourth image regions, each of which is a region of an equidistant cylindrical image, respectively, wherein in the conversion, as interpolation processing to acquire pixel values in an equidistant cylindrical image from a plurality of pixel values in a fisheye image, first interpolation processing is performed on a specific region in the third and fourth image regions, and second interpolation processing, which is simpler than the first interpolation processing, is performed on a region other than the specific region.