Segmented CPU Image Processing for GPU-Free Speed and Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing methods using Compute Unified Device Architecture (CUDA) based Graphics Processing Units (GPUs) are costly and inefficient, leading to underutilization of CPU resources and reduced accuracy when images are downscaled for processing.
Innovation Solution
A multi-threaded CPU architecture is employed to efficiently utilize CPU resources by segmenting image data across available cores, processing each segment in parallel, and combining the results for faster and accurate image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If CUDA based GPU is used for image processing, then processing speed is improved, but hardware cost increases
Solution Approach 1:
The patent segments the image into multiple regions and distributes them across multiple CPU cores for parallel processing. Each core processes a specific segment independently, then the results are combined. This segmentation approach enables efficient utilization of existing CPU resources without requiring expensive GPU hardware, thereby achieving fast processing speed while controlling hardware cost.
2Device complexity
If CPU is used for image processing, then hardware cost is reduced, but processing speed decreases due to underutilization of multi-core resources
Solution Approach 1:
The patent dynamically allocates image segments to available CPU cores based on their processing capacity and the specific requirements of the image processing task. The system monitors core utilization and adjusts segment distribution in real-time, ensuring that all multi-core resources are actively utilized rather than idle. This dynamic approach maximizes processing speed while maintaining cost-effectiveness by leveraging existing CPU hardware.
3Speed
If image is downscaled for processing, then processing speed is improved, but processing accuracy deteriorates
Solution Approach 1:
Instead of downscaling the entire image, the patent segments the original high-resolution image into multiple regions that are processed in parallel by different CPU cores. Each core works on a specific segment at full resolution, preserving detail and accuracy. The segmented approach allows the system to process large images efficiently without sacrificing quality, as each segment maintains its original resolution while the parallel processing accelerates overall computation.
Data Source
AI summary
The disclosure provides a method, a system, and a computer program product for image processing using segmentation. The system is configured to, for example, obtain image data associated with one or more sensors associated with a user device. Further the system is configured to determine count data associated with one or more cores of a processor of the user device. The image data is segmented into one or more segments based on the count data. Each of the one or more segments of the image data are processed with a corresponding core of the one or more cores of the processor of the user device. Each of the processed one or more segments of the image data are then combined to form combined processed image data.