ArticlePDF Available

A Novel Skew Estimation Approach Based on Same Height Grouping

Authors:
  • Noroff University College

Abstract and Figures

In this paper, we proposed a method to detect the rotation angle for the back side image of a container. Our technique consists of sorting the segmented characters according to the X axis, then detecting the segments belonging to the container number, after that, we divide the segments by groups having the same height, and finally the rotation angle of the image will be the average of the skewed angles related to each group. Our approach is robust, efficient and capable of handling any font and size of characters; regarding its complexity for an image having N lines and M characters, the worst CPU time usage and the worst memory usage is equal to O (NxM) while the network usage and disk usage for one image is O (1) which led –while using an old laptop- to a response time around 0.45 milliseconds to detect the rotation angle of an image when rotated from -45o to +45 o with an precision error between ±0.2o . The high accuracy and the fast response time for detecting the rotation angle of container images make our approach suitable for online OCR critical applications.
Content may be subject to copyright.
A preview of the PDF is not available
... Due to the high importance of containers in the shipping field, and in order to facilitate and accelerate trade (e.g. Customs gates, terminal operators gates, quays, …) by automatically recognizing the container number, and after presenting new techniques applied on container images for reading skewed images without image rotation [24] and for skew estimation [25], we present hereafter our proposed approach for one of the crucial phases of image binarization that we tested and returned excellent results on container images. Fig. 1 illustrates an image taken for the back side of a container. ...
... Before delving into our proposed method, it is worthy to note that neither de-skewing nor binarization of container image techniques will be described in this article because they were previously detailed in the following three papers: Reading skewed images without image rotation [11], A novel skew estimation approach based on same height grouping [12] and, Accurate, swift and noiseless image binarization [13] and this paper is a continuation of them; therefore, we will focus on our contribution to enhance the 8-neighborhood connectivity for a better ACCR results. ...
Article
Full-text available
In this paper, we will present a new variant of the 8-neighborhood connectivity; our approach remedies the segmentation problem related to scratched container code digits. Our approach is highly suitable for real-time automatic container code recognition applications because it treats many special cases, its average response time is equal to 21 milliseconds, and it improves the container code extraction and recognition by 0.89%; due to our contribution in enhancing the segmentation phase, the container code extraction accuracy reached 98.7%.
... Due to the high importance of containers in the shipping field, and in order to facilitate and accelerate trade by automatically recognizing the container number (e.g. Customs gates, terminal operators gates, quays, ), and after presenting new techniques applied on container images for reading skewed images without image rotation [24] and for skew estimation [25], we present hereafter our proposed approach for one of the crucial phases of image binarization that we tested and returned excellent results on container images. Fig. 1 illustrates an image taken for the back side of a container. ...
Article
Full-text available
In this paper, we present an accurate, swift and noiseless image binarization technique that was tested on real life back side container images. Our approach consists of transferring a colored image into grayscale, then to construct the histogram which will be divided into group of colors and after that, the foreground -that can be darker or lighter than the background- will be automatically identified and finally, the foreground boundaries will be assiduously enhanced before creating the binarized image. The average processing time of our proposed method is less than 8 milliseconds which makes it highly suitable for real life multi-user applications.
Article
Full-text available
Skewing of the scanned image is an inevitable process and its detection is an important issue for document recognition systems. The skew of the scanned document image specifies the deviation of the text lines from the horizontal or vertical axis. This paper surveys methods to detect this skew in two steps, dimension reduction and skew estimation. These methods include projection profile analysis, Hough Transform, nearest neighbor clustering, cross-correlation, piece-wise painting algorithm, piece-wise covering by parallelogram, transition counts, morphology.
Article
Full-text available
An important pre-processing task in an automated document image processing system is detection of skew in the image. Skew detection and correction is the first step in the document analysis and understanding processing steps. Correction the skewed scanned document image is very important, because it has a direct effect on the reliability and efficiency of the segmentation and feature extraction stages. It is desirable to have a fast, all-purpose skew detector. This paper proposes a novel two phase method. This two phase method reduces the skewed data before applying the skew detection and correction methods. A combination of the two phase method and the Hough Transformation can quickly and accurately detect skew on a wide variety of document images. Results of a test of the new algorithm and a comparison against the simple Hough transformation algorithm on 100 pages show the new algorithm to be much faster.
Conference Paper
Full-text available
We propose a hierarchical segmentation algorithm that starts with a very fine over segmentation and gradually merges regions using a cascade of boundary classifiers. This approach allows the weights of region and boundary features to adapt to the segmentation scale at which they are applied. The stages of the cascade are trained sequentially, with asymetric loss to maximize boundary recall. On six segmentation data sets, our algorithm achieves best performance under most region-quality measures, and does it with fewer segments than the prior work. Our algorithm is also highly competitive in a dense over segmentation (super pixel) regime under boundary-based measures.
Conference Paper
Full-text available
Feature extraction, coding and pooling, are important com-ponents on many contemporary object recognition paradigms. In this paper we explore novel pooling techniques that encode the second-order statistics of local descriptors inside a region. To achieve this effect, we introduce multiplicative second-order analogues of average and max-pooling that together with appropriate non-linearities lead to state-of-the-art performance on free-form region recognition, without any type of feature coding. Instead of coding, we found that enriching local descrip-tors with additional image information leads to large performance gains, especially in conjunction with the proposed pooling methodology. We show that second-order pooling over free-form regions produces results superior to those of the winning systems in the Pascal VOC 2011 seman-tic segmentation challenge, with models that are 20,000 times faster.
Article
Full-text available
In this manuscript the algorithm for identification of the initial skew rate for printed text is presented. Proposed algorithm creates rectangular hull around all text characters. Combining nearby rectangular hulls form objects. After applying mathematical morphology on it, the biggest object is characterized as well as selected. Rectangular hull gravity center forms reference points on these objects used as a base for calculation ie estimation of the initial skew rate. Using the least square method, initial skew rate is calculated. Comparative analysis of the origin and estimated skew rate is presented as well as discussed. Algorithm is examined with a number of printed text examples. Proposed algorithm showed robustness for skewness of printed text in the wide range.
Conference Paper
The detection and correction of document skew is one of the most important document image analysis steps. The ICDAR2013 Document Image Skew Estimation Contest (DISEC'13) is the first contest which is dedicated to record recent advances in the field of skew estimation using well established evaluation performance measures on a variety of printed document images. The benchmarking dataset that is used contains 1550 images that were obtained from various sources such as newspapers, scientific books and dictionaries. The document images contain figures, tables, diagrams, architectural plans, electrical circuits and they are written in various languages such as English, Chinese and Greek. This paper describes the details of the contest including the evaluation measures used as well as the performance of the twelve methods submitted by ten different groups along with a short description of each method.
Article
A novel Pattern Accumulated Vector(PAV) method implemented by CAM-based architecture has been proposed for character recognition in license plate recognition(LPR). Instead of comparing all the pattern blocks, the PAV method adopts the principal pattern blocks as feature vectors so that the computation complexity is highly reduced. Furthermore, the PAV method adopts the extended templates for the character recognition and is proved to have higher recognition rate when dealing with license plates consisting inclined characters. Finally, the CAM architecture is used to further reduce the processing time and makes the PAV method more efficient for real-time LPR systems.
Article
This paper makes a study on fast skew angle detection for document images. Given a document image and return the skew angle. First this algorithm computes the horizontal gradient image by the input image and then merges the characters area in the text line block, to get the valid text line by non-text area filtering, at last, statistic the text line characteristics to estimate the input image skew angle. This paper gives the processing of the algorithm. The experiment results indicate that this algorithm is accurate and effective, also with robustness. The algorithm can be applied in real- time processing system, by the more this algorithm can be applied in automatic de-skew for picture and text mixed images, layout analysis, picture and text segmentation and content-based retrieval.