Catálogo de publicaciones - libros

Discrimination between graphical drawings is a difficult problem. It can be considered at different levels according to the applications, details can be observed or more globally what could be called the style. Here we are concerned with a global view of initial letters extracted from early renaissance printed documents. We are going to present a new method to index and classify ornamental letters in ancient books. We show how the Zipf law, originally used in mono-dimensional domains can be adapted to the image domain. We use it as a model to characterize the distribution of patterns occurring in these special drawings that are initial letters. Based on this model some new features are extracted and we show their efficiency for style discrimination.

A new biometric technology-recognition ocular fundus based on shape of blood vessel skeleton-is addressed in this paper. The gray scale image of ocular fundus is utilized to extract the skeletons of its blood vessels. The cross points on the skeletons are used to match two fundus images. Experiments show high recognition rate, low recognition rejection rate as well as good universality, exclusiveness and stability of this method.

In this paper, a novel adaptive noise reduction method for engineering drawings is proposed based on the assessment of both primitives and noise. Unlike the current approaches, our method takes into account the special features of engineering drawings and assesses the characteristics of primitives and noise such that adaptive procedures and parameters are applied for noise reduction. For this purpose, we first analyze and categorize various types of noise in engineering drawings. The algorithms for average linewidth assessment, noise distribution assessment and noise level assessment are then proposed. These three assessments are combined to describe the features of the noise of each individual engineering drawing. Finally, median filters and morphological filters, which can adjust their template size and structural element adaptively according to different noise level and type, are used for adaptive noise reduction. Preliminary experimental results show that our approach is effective for noise reduction of engineering drawings.

In this paper, a method for automatically indexing the contents to reduce the effort that used to be required for input paper information and constructing index is sought. Various contents formats for journals, which have different features from those for general documents, are described. The principal elements that we want to represent are titles, authors, and pages for each paper. Thus, the three principal elements are modeled according to the order of their arrangement, and then their features are generalized. The content analysis system is then implemented based on the suggested modeling method. The content analysis system, implemented for verifying the suggested method, gets its input in the form containing more than 300 dpi gray scale image and analyze structural features of the contents. It classifies titles, authors and pages using efficient projection method. The definition of each item is classified according to regions, and then is extracted automatically as index information. It also helps to recognize characters region by region. The experimental result is obtained by applying to some of the suggested 6 models, and the system shows 97.3% success rate for various journals.

This paper proposes a character region extraction method and picture separation used for passports by adopting a preprocessing phase for passport recognition system. Character regions required in recognition make black pixel and remainder of the passport regions make white pixel in the detected character spaces. This method uses MRZ sub-region in order to automatically decide the threshold value of the binary image and this value is applied to the other character regions. Tthis method also executes horizontal and vertical histogram projection in order to remove picture region of the binary image. After the region detection of the picture area, the image part of the passport is stored in the database for face images. The remainder of the passport is composed of characters. The extraction of the picture area shows 100% of extraction ratio and the extraction of the characters for the recognition shows 96% of extraction ratio on ten different passports. From the obtained information, crosscheck process of MRZ information and field information of passport is implemented.

A method based on statistical characteristics and color space consistent with human visual perception for pixels classification is brought forward in this paper. In the airline coupon color design, we use colors to distinguish different object, the idea is embodied in this method. The marked characteristics suitable for object pixels classification have been found by analysis the statistic characteristics of all sorts of pixels. The experiments have proved that this method is simpler, more efficacious and can support data analysis for the whole coupon project.

This paper proposes a new system which supports origami creators who have no special knowledge about origami creation to create their unique works easily in 3-D virtual space. Moreover, 2-D diagrams or 3-D animation are automatically made for describing the folding processes so that people can re-build these works. Users can decide folding operations and create works by an interactive interface. For easy creation, two methods are proposed. One is a method for representing overlapping-faces of 3-D virtual origami in order to support users’ recognition of origami’s conformation. As a result, users can input information about folding operations easily and correctly. The other one is a method for deriving halfway folding processes according to users’ intents. Even if users have rough images about shapes of origami works, they may not be able to start creating an origami model as their imagination. Namely, the system shows folding processes from square to basic forms until they can start do it by themselves. We expect that the common people will create and publish their unique works and more people will enjoy origami.

A database is only usefull if it is associated a set of procedures allowing to retrieve relevant elements for the users’ needs. A lot of IR techniques have been developed for automatic indexing and retrieval in document databases. Most of these use indexes depending on the textual content of documents, and very few are able to handle graphical or image content without human annotation.

This paper describes an approach similar to the bag of words technique for automatic indexing of graphical document image databases and different ways to consequently query these databases. In an unsupervised manner, this approach proposes a set of automatically discovered symbols that can be combined with logical operators to build queries.

We present a system that recognizes tables in archival documents. Many works were carried out on table recognition but very few on tables of historical documents. These are difficult to analyze because they are often damaged due to their age and conservation. Therefore we have to introduce knowledge to compensate for missing information and noise in these documents. As there is a very important number of documents of a same type, the cost is not significant to introduce this explicit knowledge. We also want to minimalize the cost to adapt the system for a given document type. The precision of the knowledge given by the user is dependent on the quality of the document. The more the document is damaged, the more the specification has to be precise. We will show in this article how an external minimal knowledge can be sufficient for an efficient recognition system for tables in archival documents.

We present an approach for recognising mathematical texts using an extensive symbol database and a novel recognition algorithm. The process consists essentially of three steps: Recognising the individual characters in a mathematical text by relating them to glyphs in the database of symbols, analysing the recognised glyphs to determine the closest corresponding symbol, and reassembling the text by putting the appropriate commands at their corresponding positions of the original text inside a picture environment. The recogniser itself is based on a novel variation on the application of geometric moment invariants. The working system is implemented in Java.