Image Binarization via Edge Luminance Histograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately binarizing documents with varying background luminance, leading to difficulties in distinguishing characters from the background, especially when the luminance difference between characters and background is small or when the background luminance varies significantly across the document.
Innovation Solution
An image processing apparatus and method that extracts edge pixels, generates histograms based on luminance values and neighboring pixel luminance, calculates a static threshold, and uses this threshold for binarization to differentiate characters from the background effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single binarization threshold is used for the entire document image, then the processing is simple and fast, but the binarization accuracy deteriorates when background luminance varies across different regions
Solution Approach 1:
The document image is divided into multiple sub-regions based on edge detection results. Each sub-region is then processed independently with its own binarization threshold, allowing the system to adapt to local luminance variations while maintaining overall processing efficiency.
Solution Approach 2:
Different binarization thresholds are applied to different sub-regions of the document image based on their local luminance characteristics. This local adaptation ensures high binarization accuracy in each region without requiring complex global processing.
2Measurement precision
If edge detection is used to extract sub-regions for local binarization, then binarization accuracy improves in documents with varying background luminance, but character extraction fails when luminance difference between characters and background is too small
Solution Approach 1:
A luminance histogram analysis is introduced as an intermediary step to determine appropriate binarization thresholds. The histogram provides statistical information about luminance distribution, enabling the selection of thresholds that work reliably even when character-background luminance differences are minimal.
Solution Approach 2:
The binarization threshold parameter is dynamically adjusted based on luminance histogram analysis. By changing the threshold parameter according to the actual luminance distribution in each sub-region, the system maintains reliable character extraction across varying document conditions.
3Measurement precision
If the binarization threshold is set based on histogram peaks, then the method works well for documents with distinct background regions, but fails when high-density and low-density regions cannot be clearly separated in the histogram
Solution Approach 1:
The binarization threshold determination is made dynamic and adaptive rather than static. The system analyzes the luminance histogram characteristics of each sub-region and automatically adjusts the threshold selection strategy based on the observed density distribution patterns, making it versatile across different document types.
Solution Approach 2:
Luminance histogram analysis is performed in advance before final binarization. This preliminary action provides insight into the luminance distribution characteristics, enabling the system to prepare appropriate threshold values that adapt to the specific document being processed.
Data Source
AI summary
There are provided an image processing apparatus, image processing method and a computer-readable, non-transitory medium that can binarize the input image so that the characters can be differentiated with high accuracy from the background area. The image processing apparatus includes an edge pixel extractor for extracting edge pixels from an input image, a first histogram generator for generating a first histogram based on a luminance value of each of the edge pixels, a second histogram generator for generating a second histogram based on a minimum luminance value among the luminance values of pixels neighboring each of the edge pixels, a static threshold calculator for obtaining a static threshold based on the first histogram and the second histogram, and a binarization unit for binarizing the input image by using the static threshold.