Hybrid Optical Sensor Demosaicing for Resolution Enhancement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hybrid optical sensor arrays generate low-resolution images due to interspersion of visible and infrared pixels, leading to reduced fidelity, image artifacts, and edge preservation issues, which affect machine vision applications such as head tracking and spatial mapping.
Innovation Solution
Demosaicing filters are applied to images from hybrid optical sensor arrays to increase image resolution and preserve edges by interpolating missing pixel values from one subset to another, ensuring full-resolution images are produced.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If visible and infrared pixels are interspersed in a hybrid optical sensor array, then the sensor can capture both visible and infrared light simultaneously, but the image resolution is reduced
Solution Approach 1:
The patent divides the image processing task into separate processing paths for visible light pixels and infrared pixels. Each subset of pixels is processed independently through dedicated demosaicing filters, allowing resolution enhancement without compromising the hybrid spectral capture capability. This segmentation resolves the contradiction by treating each pixel type's data separately while maintaining their spatial relationship in the final composite image.
Solution Approach 2:
The patent applies demosaicing filters that operate in the spatial domain to interpolate missing pixel values, effectively adding information in the spatial dimension. By using filters that consider neighboring pixels and apply interpolation algorithms, the system recovers high-resolution information from the downsampled hybrid sensor data, transforming low-resolution interspersed pixel data into high-resolution separate images.
2Manufacturing precision
If demosaicing filters are applied to interspersed pixel data, then image resolution is enhanced, but image artifacts may be introduced
Solution Approach 1:
The patent applies different demosaicing filter characteristics to different spatial regions and pixel types. Visible light pixels and infrared pixels receive tailored filtering treatments based on their local neighborhood characteristics and spectral properties. This local adaptation allows the system to enhance resolution while preserving edge integrity and minimizing artifacts in different regions of the image.
Solution Approach 2:
The patent incorporates feedback mechanisms where the processed visible light image and infrared image are used to guide and refine each other's demosaicing process. Edge detection from one modality informs the interpolation process of the other, creating a feedback loop that suppresses artifacts while maintaining resolution enhancement. The filtered outputs are continuously refined based on comparative analysis of both spectral channels.
3Manufacturing precision
If separate processing is applied to visible and infrared pixel subsets, then full-resolution images can be generated, but processing complexity increases
Solution Approach 1:
The patent implements a universal demosaicing framework that handles both visible light and infrared pixels through a common architectural structure. The same basic filtering and interpolation algorithms are applied to both pixel types, with parameters adapted rather than completely separate implementations. This universal approach reduces processing complexity compared to entirely separate processing pipelines while still achieving full-resolution output for both modalities.
Data Source
AI summary
A hybrid optical sensor array includes a first and second set of pixels having differing spectral sensitivities. A first set of data for a scene is captured by the first set of pixels at a first resolution, and a second set of data for the scene is captured by a second set of pixels at a second resolution. The first set of data is demosaiced based on at least the second set of data. A third set of data for the scene is output at a third resolution, greater than the first resolution. This allows high resolution matching images to be produced without perspective or timing discrepancies inherent in multi-camera machine vision systems.