Image Fusion Imaging for Real-Time Ultra-Large Pixel Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing imaging technologies for ultra-large-scale pixels face challenges in capturing and transmitting images in real-time due to high data volume and slow reading speeds, leading to poor image quality and inability to capture video in real-time.
Innovation Solution
A method involving image collection and fusion technology that combines high-resolution low-bit-width differential images with low-resolution high-bit-width images using image fusion algorithms, including convolutional neural networks, to efficiently reconstruct ultra-large-scale pixel images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-chip ultra-large-scale pixel imaging is used to capture ultra-large-scale pixel images, then the imaging resolution is improved, but the reading speed becomes slow and data volume increases
Solution Approach 1:
The imaging system is divided into multiple independent pixel chips (e.g., 10x10 array of 5MP chips) that can be read out simultaneously through parallel data buses. Each chip independently captures and processes image data, enabling ultra-large-scale imaging (e.g., 500MP) while maintaining high reading speeds through parallelization. This segmentation resolves the contradiction by distributing the imaging load across multiple units rather than using a single large chip.
2Measurement precision
If single-chip ultra-large-scale pixel imaging is used to capture ultra-large-scale pixel images, then the imaging resolution is improved, but the data volume increases making real-time transmission impossible
Solution Approach 1:
The system extracts only the essential image data from each pixel chip and transmits it through parallel data buses to a central processing unit. By extracting and processing data in a distributed manner rather than transmitting complete high-resolution images from a single chip, the system reduces transmission bandwidth requirements while maintaining ultra-large-scale imaging capability. The parallel architecture allows data to be taken out and processed incrementally across multiple channels.
3Measurement precision
If traditional interpolation algorithm or image restoration algorithm based on neural network is used to magnify low-resolution image, then the resolution is improved, but information loss occurs leading to poor imaging effect
Solution Approach 1:
The system performs preliminary high-resolution data capture at the sensor level through parallel pixel arrays before any image processing or magnification occurs. By capturing ultra-large-scale pixel data directly from the sensor array in its native high-resolution form, the system preserves all original information without subsequent interpolation or restoration operations. This preliminary action of capturing maximum detail at the source eliminates information loss that would otherwise occur during resolution enhancement.
Data Source
AI summary
Provided are an imaging method and apparatus, and a medium. The method includes: for a target field of view, performing photographing and quantification at a first resolution, to obtain a first image that has a first bit width; for the target field of view, performing photographing and differential processing at a second resolution, to obtain a second image that has a second bit width, wherein the differential processing includes: for a pixel point obtained by photographing at the second resolution, quantifying the difference between the pixel point and a neighboring pixel point of the pixel point, to obtain a quantified difference a value of a corresponding pixel point in the second image; and fusing the first image with the second image, to obtain a third image, wherein the first resolution is lower than the second resolution, and the first bit width is higher than the second bit width.