Catálogo de publicaciones - libros

An orthogonal polygonal line is a line consisting of adjacent straight segments having only two directions orthogonal to each other. Because of noise and vectorization errors, the result of vectorization of such a line may differ from an orthogonal polygonal line. This paper contains the description of an optimal method for the restoration of orthogonal polygonal lines. It is based on the method of restoration of arbitrary ground truth lines from the paper [1]. Specificity of the algorithm suggested in the paper consists of filtering vectorization errors using a priori information about orthogonality of the ground truth contour. The suggested algorithm guarantees that obtained polygonal lines will be orthogonal and have minimal deviations from the ground truth line. The algorithm has a low computational complexity and can be used for restoration of orthogonal polygonal lines with many vertices. It was developed for a raster-to-vector conversion system ArcScan for ArcGIS and can be used for interactive vectorization of orthogonal polygonal lines.

With a great amount of textual information are available on the Internet and corporate intranets, it has become a necessary to categorize large documents. As we known, text classification problem is representative multiclass problem. This paper describes a framework, which we call Strong-to-Weak- to-Strong (SWS). It transforms a “strong” learning algorithm to a “weak” algorithm by decreasing its iterative numbers of optimization while preserving its other characteristics like geometric properties and then makes use of the kernel trick for “weak” algorithms to work in high dimensional spaces, finally improves the performances of text classification. We analyzed the particular properties of learning with text and identified why this approach is appropriate for this task. Empirical results show that our approach is competitive with the other methods.

This paper presents a document restoration technique that is able to flatten curled document images captured through a digital camera. The proposed method corrects camera images of documents through image partition, which divides distorted text lines into multiple small patches based on the identified vertical stroke boundary (VSB) and the fitted x-line and baseline of text lines. Target rectangles are then constructed through the exploitation of the characters enclosed within the partitioned image patches. With the constructed target rectangles and the partitioned image patches, global geometric distortion is finally removed through the local rectification of partitioned image patches one by one. Experimental results show that the proposed technique is fast, accurate, and easy for implementation.

A multi-classifier formed by specialised classifiers for noise produced by an image is shown in this work. A study has been carried out in the case of cut images, where tree cases of specialization are considered. Classifiers based on neighbourhood criteria are used, the zoning global feature and the Euclidean distance too. Furthermore, the paper explains a modification of the Euclidean distance for classifying cut digits. The experiments have been carried out with images of typewritten digits, taken from real forms. Trying to obtain a strong database to support the experiments, we have cut images deliberately. The recognition rate improves from 84.6% to 97.70%, but whether the system provides information about the disturbance of the image, it can achieve a 98.45%.

Digitization process of various printed documents involves generating texts by an OCR system for different applications including full-text retrieval and document organizations. However, OCR-generated texts have errors as per present OCR technology. Moreover, previous studies have revealed that as OCR accuracy decreases the classification performance also decreases. The reason for this is the use of absolute word frequency as feature vector. Representing OCR texts using absolute word frequency has limitations such as dependency on text length and word recognition rate consequently lower classification performance due to higher within-class variances. We describe feature transformation techniques which do not have such limitations and present improved experimental results from all used classifiers.

In this paper we propose a new approach to find symbols in graphical documents. The method is based on a representation of the document in chain points extracted from the skeleton. We merge successively these chain points into a dendrogram framework and according to a measure of density. From the dendrogram, we extract potential symbols which can be recognized after.

Historical sound documents are of high importance for our cultural heritage. The sound of phonographic records is usually extracted by a stylus following the groove, but many old records are in such bad shape that no mechanical contact is possible. The only way to read them is by a contactless reading system. A phonographic document analysis system was developed using an optical technique to retrieve the sound from old records. The process is straightforward: we take a picture of each side of the disc using a dedicated analog camera, we store the film as our working copy, and when needed, we scan the film and process the image in order to extract the sound. In this paper, we analyze the imaging issues and present the algorithm for extracting the groove position and therefore the sound of the records.

Camera-based document image retrieval is a task of searching document images from the database based on query images captured using digital cameras. For this task, it is required to solve the problem of “perspective distortion” of images,as well as to establish a way of matching document images efficiently. To solve these problems we have proposed a method called Locally Likely Arrangement Hashing (LLAH) which is characterized by both the use of a perspective invariant to cope with the distortion and the efficiency: LLAH only requires O ( N ) time where N is the number of feature points that describe the query image. In this paper, we introduce into LLAH an affine invariant instead of the perspective invariant so as to improve its adjustability. Experimental results show that the use of the affine invariant enables us to improve either the accuracy from 96.2% to 97.8%, or the retrieval time from 112 msec./query to 75 msec./query by selecting parameters of processing.

As the spread of digital videos, digital cameras, and camera phones, lots of researches are reported about degraded character recognition. It is found that while the grayscale-based classifier is powerful for degraded character, the performance for clear character is not so good as binary-based classifier. In this paper, a dynamic classifier selection method is proposed to combine the two classifiers based on an estimation of the degradation level and the recognition reliability of the input character images. Experimental results show that the proposed method can achieve better recognition performance than the two individual ones.

In this paper, we present a method for the automatic extraction of numerical fields (ZIP codes, phone numbers, etc.) from incoming mail documents. The approach is based on a segmentation-driven recognition that aims at locating isolated and touching digits among the textual information. A syntactical analysis is then performed on each line of text in order to filter the sequences that respect a particular syntax (number of digits, presence of separators) known by the system. We evaluate the performance of our system by means of the recall precision trade-off on a real incoming mail document database.