Unsupervised refinement of color and stroke features for text binarization

Color and strokes are the salient features of text regions in an image. In this work, we use both these features as cues, and introduce a novel energy function to formulate the text binarization problem. The minimum of this energy function corresponds to the optimal binarization. We minimize the energy function with an iterative graph cut-based algorithm. Our model is robust to variations in foreground and background as we learn Gaussian mixture models for color and strokes in each iteration of the graph cut. We show results on word images from the challenging ICDAR 2003/2011, born-digital image and street view text datasets, as well as full scene images containing text from ICDAR 2013 datasets, and compare our performance with state-of-the-art methods. Our approach shows significant improvements in performance under a variety of performance measures commonly used to assess text binarization schemes. In addition, our method adapts to diverse document images, like text in videos, handwritten text images.     

Document image binarization using background estimation and stroke edges

Document images often suffer from different types of degradation that renders the document image binarization a challenging task. This paper presents a document image binarization technique that segments the text from badly degraded document images accurately. The proposed technique is based on the observations that the text documents usually have a document background of the uniform color and texture and the document text within it has a different intensity level compared with the surrounding document background. Given a document image, the proposed technique first estimates a document background surface through an iterative polynomial smoothing procedure. Different types of document degradation are then compensated by using the estimated document background surface. The text stroke edge is further detected from the compensated document image by using L1-norm image gradient. Finally, the document text is segmented by a local threshold that is estimated based on the detected text stroke edges. The proposed technique was submitted to the recent document image binarization contest (DIBCO) held under the framework of ICDAR 2009 and has achieved the top performance among 43 algorithms that are submitted from 35 international research groups.     

一种基于迭代阈值法的身份证图像二值化算法研究

针对身份证扫描图像受激光防伪阴影网格线影响较大的特点,对图像二值化的方法进行了深入分析和研究．提出了一种基于像素邻域特征的迭代阈值方法。该算法模型简单、易于实现．能滤除噪声、使字符笔划清晰地从背景中分割出来．获得了较好的二值化效果。     

基于局部对比度和相位保持降噪的古籍图像二值化算法

图像二值化算法通过消除文档背景噪声将文本与背景分割开。针对古籍图像提出一种基于局部对比度和相位保持降噪的古籍图像二值化算法。根据归一化局部最大值最小值来构造古籍图像局部对比度,同时对古籍图像进行相位保持降噪。将局部对比度图像和降噪图像相结合来识别文本笔划像素。通过局部窗口内所检测的文本笔划像素估计局部阈值从而计算古籍背景修复模板。用图像修复算法和形态学闭操作来估计古籍背景。用所估计背景来增强古籍图像,采用Howe算法对增强后的古籍图像进行二值化求得最终结果。该算法在DIBCO2016、DIBCO2017和DIBCO2018数据集中的实验结果均优于其他二值化算法。     

Adaptive binarization method for degraded document images based on surface contrast variation

Document binarization is an important technique in document image analysis and recognition. Generally, binarization methods are ineffective for degraded images. Several binarization methods have been proposed; however, none of them are effective for historical and degraded document images. In this paper, a new binarization method is proposed for degraded document images. The proposed method based on the variance between pixel contrast, it consists of four stages: pre-processing, geometrical feature extraction, feature selection, and post-processing. The proposed method was evaluated based on several visual and statistical experiments. The experiments were conducted using five International Document Image Binarization Contest benchmark datasets specialized for binarization testing. The results compared with five adaptive binarization methods: Niblack, Sauvola thresholding, Sauvola compound algorithm, NICK, and Bataineh. The results show that the proposed method performs better than other methods in all binarization cases.     

加权KNN的图文数据融合分类

目的 图文数据在不同应用场景下的最佳分类方法各不相同,而现有语义级融合算法大多适用于图文数据分类方法相同的情况,若将其应用于不同分类方法时由于分类决策基准不统一导致分类结果不理想,大幅降低了融合分类性能。针对这一问题,提出基于加权KNN的融合分类方法。方法 首先,分别利用softmax多分类器和多分类支持向量机(SVM)实现图像和文本分类,同时利用训练数据集各类别分类精确度加权后的图像和文本正确判别实例的分类决策值分别构建图像和文本KNN模型;再分别利用其对测试实例的图像和文本分类决策值进行预测,通过最邻近k个实例属于各类别的数目确定测试实例的分类概率,统一图像和文本的分类决策基准;最后利用训练数据集中图像和文本分类正确的数目确定测试实例中图像和文本分类概率的融合系数,实现统一分类决策基准下的图文数据融合。结果 在Attribute Discovery数据集的图像文本对上进行实验,并与基准方法进行比较,实验结果表明,本文融合算法的分类精确度高于图像和文本各自的分类精确度,且平均分类精确度相比基准方法提高了4.45%;此外,本文算法对图文信息的平均整合能力相比基准方法提高了4.19%。结论 本文算法将图像和文本不同分类方法的分类决策基准统一化,实现了图文数据的有效融合,具有较强的信息整合能力和较好的融合分类性能。     

Integrating computer vision with web‐based knowledge for medical diagnostic assistance

Abstract: The analysis of medical documents necessitates context recognition for diverse purposes such as classification, performance analysis and decision making. Traditional methods of context recognition have focused on the textual part of documents. Images, however, provide a rich source of information that can support the context recognition process. A method is proposed for integrating computer vision in context recognition using the web as a knowledge base. The method is implemented on medical case studies to determine the main symptoms or achieve possible diagnoses. In experiments the method for integrating computer vision in context recognition achieves better results than term frequency and inverse document frequency and only context recognition. The proposed method can serve as a basis for an image and text based decision support system to assist the physician in reviewing medical records.     

A binarization method with learning-built rules for document images produced by cameras

In this paper, we propose a novel binarization method for document images produced by cameras. Such images often have varying degrees of brightness and require more careful treatment than merely applying a statistical method to obtain a threshold value. To resolve the problem, the proposed method divides an image into several regions and decides how to binarize each region. The decision rules are derived from a learning process that takes training images as input. Tests on images produced under normal and inadequate illumination conditions show that our method yields better visual quality and better OCR performance than three global binarization methods and four locally adaptive binarization methods.     

Parameter-free based two-stage method for binarizing degraded document images

Binarization plays an important role in document image processing, especially in degraded documents. For degraded document images, adaptive binarization methods often incorporate local information to determine the binarization threshold for each individual pixel in the document image. We propose a two-stage parameter-free window-based method to binarize the degraded document images. In the first stage, an incremental scheme is used to determine a proper window size beyond which no substantial increase in the local variation of pixel intensities is observed. In the second stage, based on the determined window size, a noise-suppressing scheme delivers the final binarized image by contrasting two binarized images which are produced by two adaptive thresholding schemes which incorporate the local mean gray and gradient values. Empirical results demonstrate that the proposed method is competitive when compared to the existing adaptive binarization methods and achieves better performance in precision, accuracy, and F-measure.     

一种改进的针对退化文本图像的二值化方法

在经典的Niblack方法的基础上提出了一种改进的针对退化文本图像的二值化方法,该方法仅在文本区域周围较小范围内进行局部阈值计算,在大大减少运算量的同时,克服了Niblack方法容易产生大量背景噪声的缺点,与另外一种同样基于Niblack的Sauvola方法相比较,对于低对比度的退化文本图像有更好的适应性。