Multi-SoC Image Processing via Data Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current system-on-chip (SoC) technologies are not cost-effective for achieving high arithmetic capabilities, as using high-end SoCs in low-end products is costly, and existing multi-core and multi-cluster technologies do not effectively leverage the cooperation of multiple SoCs for enhanced image processing performance.

Innovation Solution

An image processing device comprising multiple SoCs with a data splitter to divide input image data into parts, processed by each SoC, and then recombined, exceeding the processing capabilities of individual SoCs, with external circuits for data transmission and synchronization, allowing for higher performance without exceeding individual processing limits.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple SoCs with different arithmetic capabilities are used to achieve high processing performance, then the image processing capability is improved, but the total research, development, and manufacturing cost increases

Engineering Contradiction:
Improveimage processing capabilityVSAvoidresearch, development, and manufacturing cost
Core Design Contradiction:
ProductivityVSEase of manufacture

Solution Approach 1:

The patent divides the image processing task into multiple segments, with each SoC processing a specific portion of the input image data. The data splitter divides the input image data into multiple parts, and each image processing circuit processes one part independently. This segmentation allows multiple identical or similar low-cost SoCs to work together to achieve high processing performance without the need for expensive high-end SoCs.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent uses multiple copies of the same or similar SoC design rather than one high-end SoC. Each SoC is an identical or similar low-arithmetic capability unit that processes a portion of the data. By copying the same proven low-cost design multiple times, the system achieves high aggregate performance while avoiding the high R&D and manufacturing costs of developing and producing high-end SoCs.

Inventive Principle:
Principle #26Copying

2Productivity

If a single high-arithmetic capability SoC is used, then the image processing performance is improved, but the cost-effectiveness deteriorates

Engineering Contradiction:
Improveimage processing performanceVSAvoidcost-effectiveness
Core Design Contradiction:
ProductivityVSEase of manufacture

Solution Approach 1:

Instead of using one high-performance SoC, the patent segments the processing workload across multiple lower-performance SoCs. Each SoC handles a specific portion of the image data, and their combined output achieves the desired high performance. This segmentation makes the system more cost-effective by using affordable components in quantity rather than expensive components in singularity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges the processing outputs of multiple identical or similar SoCs to achieve high image processing performance. The data splitter divides the input data, and after each SoC processes its assigned portion, the results are combined to form the complete processed image. This merging approach allows the system to achieve performance comparable to or exceeding that of a single high-end SoC while maintaining cost-effectiveness.

Inventive Principle:
Principle #5Merging (Combining)

3Productivity

If existing multi-core and multi-cluster technologies are used, then the processing capability is improved, but the ability to leverage cooperation of multiple SoCs is lost

Engineering Contradiction:
Improveprocessing capabilityVSAvoidcooperation of multiple SoCs
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent extends segmentation from within a single SoC to across multiple independent SoCs. Each SoC is a complete system with its own processor and image processing circuit, and the segmentation occurs at the system level rather than just at the core level within one chip. This allows each SoC to be an independent functional unit that can cooperate with others.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces external circuits as intermediaries that facilitate cooperation between multiple independent SoCs. These external circuits handle data distribution, synchronization, and result aggregation, enabling the SoCs to work together as a coordinated system while maintaining their independence. This intermediary layer allows versatile cooperation between multiple SoCs without requiring them to be integrated into a single chip.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS11461868B2Image processing device
Publication Date: 2022.10.04 REALTEK SEMICON CORP
  • US11461868B2 patent drawing
  • US11461868B2 patent drawing
  • US11461868B2 patent drawing

AI summary

An image processing device includes a main SoC, an performance-enhancing SoC, and an external circuit set outside any of the two SoCs. The main SoC includes: a data splitter dividing input image data into a first input part and a second input part; a first image processing circuit processing the first input part to generate a first output part; and a transmitter outputting the second input part to the performance-enhancing SoC via the external circuit. The performance-enhancing SoC includes: a receiver receiving the second input part via the external circuit; and a second image processing circuit processing the second input part to generate a second output part. The combination of the two output parts jointly determines a data amount per unit of time which exceeds the processing capability of any of the two image processing circuits. Each of the two SoCs includes a CPU, and the two CPUs cooperate, too.