Solid-State Imaging Element With Variable-Coefficient Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional solid-state imaging elements with analog circuits for convolution integration suffer from limited filter coefficients (+1, 0, and −1), leading to insufficient accuracy and slow processing speed, especially when imaging moving subjects.
Innovation Solution
A solid-state imaging element with a pixel array, integration circuits, and analog-to-digital converters that perform time-integration and analog addition of pixel signals, allowing for variable integration times based on filter coefficients and digital conversion, along with a control method that reflects filter coefficient signs in signal levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If analog circuits with limited filter coefficients (+1, 0, and −1) are used for convolution integration, then processing speed is improved, but image recognition accuracy becomes insufficient
Solution Approach 1:
The patent changes the parameter of filter coefficients from limited discrete values (+1, 0, -1) to continuous variable values by using variable integration times in the integration circuits. This allows the filter coefficients to take any value within a range, significantly improving image recognition accuracy while maintaining the analog circuit's fast processing speed advantage.
2Device complexity
If analog addition of difference is performed column by column, then circuit complexity is reduced, but processing speed becomes slow for moving subjects
Solution Approach 1:
The patent segments the image processing into two stages: first, parallel time-integration of pixel signals across multiple columns simultaneously in the integration circuits; second, sequential analog addition of the integrated signals. This segmentation allows parallel processing of multiple columns at the integration stage, dramatically improving processing speed for moving subjects while keeping the addition stage relatively simple.
3Device complexity
If filter coefficients are limited to +1, 0, and −1, then device complexity is reduced, but manufacturing precision of image recognition becomes insufficient
Solution Approach 1:
The patent introduces dynamics into the previously static filter coefficient system by making the integration time variable. This allows the filter coefficients to be dynamically adjusted to any value within a continuous range, enabling high-precision image recognition applications that require sophisticated filter coefficients without increasing device complexity.
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
Improves image recognition accuracy and processing speed by enabling broader filter coefficients and simultaneous integration of multiple pixel signals, reducing noise and power consumption.
Implementation Method 1
a plurality of integration circuits that time-integrates the pixel signal held in each of a predetermined number of pixels arranged in a vertical direction among the plurality of pixels and outputs an integration signal
Implementation Method 2
an analog-to-digital converter that analog-adds the integration signal of each of the plurality of integration circuits to convert the analog-added integration signal into a digital signal
Data Source
AI summary
Accuracy is improved in a solid-state imaging element that performs image recognition by convolution integration. In a pixel array section, a plurality of pixels each generating and holding an analog pixel signal is arranged in a two-dimensional grid pattern. A plurality of integration circuits time-integrates the pixel signal held in each of a predetermined number of pixels arranged in a vertical direction among the plurality of pixels and outputs an integration signal. An analog-to-digital converter analog-adds the integration signal of each of the plurality of integration circuits and converts the analog-added integration signal into a digital signal.