Object-Based Sharpening Filter for Webcam OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCR systems require specialized hardware, such as close-focused cameras, which are cumbersome and costly, and existing webcams or laptop cameras produce defocused images at close ranges, making them unsuitable for business card OCR without pre-processing.
Innovation Solution
A method involving object-based sharpening filters and block-wise quantization is applied to defocused images from existing webcams or laptop cameras to enhance image quality for accurate OCR, allowing the use of built-in cameras for business card recognition without additional hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized close-focused camera hardware is used, then image focus quality is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces the mechanical/optical solution (specialized close-focused camera hardware) with a software-based image processing solution. By applying deconvolution algorithms and sharpening filters to defocused images captured by standard webcams, the system achieves comparable OCR accuracy without requiring specialized camera hardware, thereby reducing device complexity and cost.
2Device complexity
If standard webcam is used, then device complexity is reduced, but image focus quality deteriorates at close range
Solution Approach 1:
The patent applies preliminary image processing actions (deconvolution, sharpening filters, contrast enhancement) to the defocused image data before OCR processing. By pre-processing the blurry images captured by standard webcams with algorithms that reverse blur effects and enhance edges, the system compensates for the poor focus quality and enables accurate character recognition.
3Adaptability or versatility
If defocused image processing is applied, then adaptability of existing cameras is improved, but processing time increases
Solution Approach 1:
The patent segments the image processing task into distinct stages: initial deconvolution to reverse blur, followed by sharpening filter application, then contrast enhancement, and finally OCR processing. This segmentation allows each processing stage to be optimized independently and enables parallel processing where applicable, reducing overall processing time while maintaining adaptability to images from various camera sources.
Data Source
AI summary
A method of pre-processing a defocused image of an object includes applying an object-based sharpening filter on the defocused image to produce a sharper image; and quantizing the sharper image using block-wise quantization. A system for generating decoded text data from alphanumeric information printed upon an object includes a camera that obtains image data of the alphanumeric information. The system also includes a pre-processor that (a) performs block-wise quantization of the image data to form conditioned image data, and (b) performs optical character recognition on the conditioned image data to generate the decoded text data.