Document Edge Detection Using GPU-CPU Task Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for text recognition in documents, such as OCR, are resource-intensive and require substantial CPU resources, making real-time or fast image recognition challenging, especially when using portable devices, and often require repeated attempts due to user error like camera shaking.
Innovation Solution
A method that splits image processing tasks between a graphics processor unit (GPU) and a central processor unit (CPU) in a portable device, using the GPU for initial filtering and edge detection and the CPU for stability analysis and storage of images, allowing for real-time document edge recognition and reducing the need for manual shutter release, thereby improving processing speed and reducing errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing is performed using a central processor unit (CPU), then comprehensive document analysis can be achieved, but processing speed decreases and real-time recognition becomes unavailable
Solution Approach 1:
The patent divides image processing into two segments: edge detection and filtering are performed by a graphics processor unit (GPU) for fast processing, while comprehensive document analysis and OCR are performed by a central processor unit (CPU) for accuracy. This segmentation allows each processor to specialize in specific tasks, achieving both speed and precision.
Solution Approach 2:
The patent introduces an intermediary processing stage using a graphics processor unit that performs preliminary edge detection and filtering before the final document analysis. This intermediary layer prepares the image data in advance, reducing the processing burden on the CPU and enabling faster overall processing while maintaining accuracy.
2Ease of operation
If a user manually operates the camera shutter for document capture, then control over the capture process is maintained, but repeated attempts are required due to camera shaking and poor image quality
Solution Approach 1:
The patent implements self-service by having the system automatically detect when a document is properly positioned and captured. The graphics processor unit continuously analyzes incoming images for document edges and stability criteria, automatically triggering storage when conditions are met, eliminating the need for manual shutter operation and repeated attempts.
Solution Approach 2:
The patent introduces real-time feedback by continuously monitoring image quality and document detection results during camera operation. When the graphics processor unit detects that a document has been properly captured meeting stability criteria, it provides feedback to store the image automatically, guiding the user without requiring manual intervention or repeated attempts.
3Speed
If fast image processing is implemented using a graphics processor unit, then real-time recognition becomes possible, but device complexity increases
Solution Approach 1:
The patent applies universality by utilizing a graphics processor unit that can perform multiple functions: edge detection, filtering, real-time document analysis, and preparation for OCR. This multi-functional approach avoids adding separate dedicated hardware for each function, achieving fast processing while minimizing the increase in device complexity.
Data Source
AI summary
A method is proposed for detecting a document in which image data are recorded by means of a camera, in which filtered picture data are determined by a first processing unit on the basis of the recorded image data, and a camera picture is stored by a second processing unit on the basis of the filtered picture data if a stability criterion is fulfilled. Also specified correspondingly are a device, computer program product and storage medium.