Image Sensor Sign-Bit Extraction for Sparse Scene Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine vision systems are limited by the speed of digital camera image capture, which restricts the rate of 3D point cloud collection due to the design of CMOS image sensors, making them impractical for commercially feasible systems, especially when dealing with sparse and compressible digital images.
Innovation Solution
A method involving an image sensor with a pixel array that produces measurements using sign-bits, compares pixel currents with sampling patterns, and combines differential measurements with pixel sampling patterns to form digital images, utilizing sparse matrix multiplication and convolution to enhance image capture and processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional CMOS image sensor architecture is used to achieve ultra-high-speed image capture, then image capture speed is improved, but device complexity, size, cost and supporting system requirements worsen
Solution Approach 1:
The patent extracts only the essential function of image capture from the complex CMOS sensor system and implements it using a simplified pixel array that directly outputs analog signals. This removes unnecessary components and processing stages, achieving high-speed capture without the complexity of conventional CMOS sensors.
Solution Approach 2:
The patent replaces the mechanical/electronic readout system of conventional CMOS sensors with an optical correlation approach. By using optical correlators and analog signal processing, the system achieves ultra-high-speed operation without the bottlenecks of digital readout electronics.
2Productivity
If conventional digital camera image capture rate is used, then measurement precision is maintained, but productivity of 3D point cloud collection worsens
Solution Approach 1:
The patent enables continuous high-speed image capture by using an analog optical processing system that can operate at rates much higher than digital cameras. The continuous analog signal flow allows for uninterrupted 3D point cloud data collection, eliminating the frame-rate limitations of digital sensors.
Solution Approach 2:
The patent changes the fundamental operating parameters from digital discrete sampling to analog continuous processing. This parameter change allows the system to operate at ultra-high speeds while maintaining measurement precision through analog optical correlation techniques.
3Measurement precision
If full digital image readout is performed, then measurement precision is improved, but loss of time and energy increase due to redundant data processing
Solution Approach 1:
The patent performs preliminary processing of the image data through optical correlation before digital conversion. By pre-processing the analog signals optically, the system extracts relevant information before the time-consuming digital readout and conversion stages, reducing overall processing time while maintaining precision.
Solution Approach 2:
The patent extracts only the relevant measurement information from the full image data through optical correlation techniques. This selective extraction avoids the need to readout and process all pixel data digitally, eliminating redundant data handling while preserving measurement accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly improves the throughput of machine vision systems by reducing redundant readout and conversion, allowing for a 16x improvement in bit-rate compression and efficient encoding of sparse images, overcoming the limitations of conventional image sensor architectures.
Implementation Method 1
each of the two or more measurements including information contained in a set of sign-bits, the producing of each measurement of the two or more measurements including (i) forming an image signal on the pixel array; and (ii) comparing accumulated pixel currents output from at least some pixels of the pixel array
Data Source
AI summary
A method includes producing two or more measurements by an image sensor having a pixel array, the measurements including information contained in a set of sign-bits, the producing of each measurement including (i) forming an image signal on the pixel array; and (ii) comparing accumulated pixel currents output from pixels of the pixel array in accordance with the image signal and a set of pixel sampling patterns to produce the set of sign-bits of the measurement; buffering at least one of the measurements to form a buffered measurement; comparing information of the buffered measurement to information of the measurements to produce a differential measurement; and combining the differential measurement with information of the set of pixel sampling patterns to produce at least a portion of one or more digital images relating to one or more of the image signals formed on the pixel array.