Cell Network Image Segmentation for Low Power Consumption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image segmentation methods struggle with real-time processing of large images due to high power consumption and increased chip area, making them unsuitable for applications like battery-driven robots and portable devices.
Innovation Solution
The proposed solution involves a digital image segmentation architecture that uses a cell network with cells in non-ignited, self-ignitable, and ignited states, calculating coupling weights between adjacent cells to identify leader cells and segment regions efficiently, incorporating the Boundary Active Only (BAO) scheme and Subdivided Image Approach (SIA) to reduce power consumption and maintain real-time processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image segmentation algorithms are implemented in software, then processing accuracy is maintained, but processing time increases and real-time processing becomes difficult
Solution Approach 1:
The patent replaces software-based image segmentation with a hardware-based cell network system. Each pixel is represented by a cell that undergoes state transitions (non-ignited, self-ignitable, ignited) based on coupling weights with adjacent cells. This hardware implementation parallelizes the segmentation process, achieving real-time processing while maintaining accuracy through the same underlying algorithmic logic.
Solution Approach 2:
The image is divided into individual pixel cells, each independently processing its state transitions. This cellular automata approach segments the processing task across numerous parallel units, enabling simultaneous computation for all pixels and achieving real-time performance.
2Productivity
If hardware acceleration is used to achieve real-time processing, then processing speed increases, but power consumption and chip area increase
Solution Approach 1:
The cell network processes images in discrete time steps with periodic state transitions. Cells alternately transition between non-ignited, self-ignitable, and ignited states in synchronized cycles. This periodic operation allows for power management and reduces average power consumption compared to continuous processing, while maintaining real-time throughput.
Solution Approach 2:
The BAO (Boundary Active Only) scheme activates computation only for cells at region boundaries during segmentation. Interior cells that have already been classified remain inactive, reducing the number of active computational units and thereby lowering power consumption while preserving segmentation accuracy at critical boundary regions.
3Productivity
If hardware acceleration is used to achieve real-time processing, then processing speed increases, but chip area increases
Solution Approach 1:
Adjacent cells share common coupling weight calculations and state transition logic. The hardware architecture merges identical computational units into standardized cell modules that can be tiled across the chip, reducing redundant circuitry and minimizing overall chip area while supporting parallel processing of large images.
Solution Approach 2:
The cell network is organized as a modular grid of identical cell units, each handling a single pixel. This segmentation into reusable modules allows for scalable design where the same small cell circuit is replicated N times for an N-pixel image, optimizing chip area utilization through regularity and reducing interconnect complexity.
4Area of stationary object
If large-scale images are processed, then image detail and coverage improve, but power consumption and processing time increase
Solution Approach 1:
The BAO scheme identifies and activates only boundary cells that require processing, leaving interior cells inactive. For large images, this dramatically reduces the proportion of active cells, thereby reducing power consumption and processing time while maintaining complete image coverage and detail through boundary-driven region propagation.
Solution Approach 2:
The cell network processes large images in periodic cycles of state transitions rather than requiring all cells to compute simultaneously. This temporal segmentation of computation allows large images to be processed with reduced peak power consumption and enables progressive refinement of segmentation regions across multiple time steps.
Data Source
AI summary
In a boundary active only scheme proposed by the present invention, only a cell in a boundary of region growth is brought into an active mode, and the other cells are brought into a standby mode. The respective cells perform state transition in parallel, and decision of the state transition performed for each clock cycle is not performed in a case where any of the three conditions that none of the adjacent cells is ignited, the cell itself is already ignited, and the cell already belongs to a certain divided region is satisfied. Therefore the number of simultaneously operating cells and that of coupling weight registers are minimized, and control is automatically executed to reduce power consumption.