Theodosios Pavlidis

Page segmentation and classification

Abstract Page segmentation is the process by which a scanned page is divided into columns and blocks which are then classified as halftones, graphics, or text. Past techniques have used the fact that such parts form right rectangles for most printed material. This property is not true when the page is tilted, and the heuristics based on it fail in such cases unless a rather expensive tilt angle estimation is performed. We describe a class of techniques based on smeared run length codes that divide a page into gray and nearly white parts. Segmentation is then performed by finding connected components either by the gray elements or of the white, the latter forming white streams that partition a page into blocks of printed material. Such techniques appear quite robust in the presence of severe tilt (even greater than 10 °) and are also quite fast (about a second a page on a SPARC station for gray element aggregation). Further classification into text or halftones is based mostly on properties of the across scanlines correlation. For text correlation of adjacent scanlines tends to be quite high, but then it drops rapidly. For halftones, the correlation of adjacent scanlines is usually well below that for text, but it does not change much with distance.

Discrimination of characters by a multi-stage recognition process

Abstract This paper proposes a hierarchical character recognition scheme. The recognition process is divided into several stages. An aggressive graph-matching procedure first attempts to recognize various broad character classes. It then triggers a set of highly specific routines suitable for the recognition of particular characters or classes of characters. Two techniques are employed in dealing with the ambiguity of interpretation: one technique is to seek further constraints in the way character images are interpreted; the other is to use additional measurements to curtail the ambiguity of subsequent interpretations.

Feature analysis using line sweep thinning algorithm

We propose a new thinning algorithm based on line sweep operation. Assuming that the contour of the figure to be thinned has been approximated by polygons, the events are then the vertices of the polygons, and the line sweep algorithm searches for pairs of edges lying within each slab. The pairing of edges is useful for detecting both regular and intersection regions. The regular regions can be found at the sites where pairings between edges exist. Intersection regions are those where such relations would cease to exist. A salient feature of our approach is that it finds simultaneously the set of regular regions that attach to the same intersection region. Such a set is thus called an intersection set. The output of our algorithm consists of skeletons as well as intersection sets, both can be used as features for subsequent character recognition. Moreover, the line sweep thinning algorithm is efficient in computation as compared with a pixel-based thinning algorithm which outputs skeletons only.In this article, we propose a new thinning algorithm based on line sweep operation. A line sweep is a process where the plane figure is divided into parallel slabs by lines passing through certain ...

A note on the trade-off between sampling and quantization in signal processing

Abstract We discuss the trade-off between sampling and quantization in signal processing for the purpose of minimizing the error of the reconstructed signal subject to the constraint that the digitized signal fits in a given amount of memory. For signals with different regularities, we estimate the intrinsic errors from finite sampling and quantization, and determine the sampling and quantization resolutions.

A Hierarchical Approach to Efficient Curvilinear Object Searching

Curvilinear object searching is a common problem encountered in pattern recognition and information retrieval. How to improve the efficiency of searching is the major concern, especially when the data set is large. In this paper we propose a hierarchical approach, where high-level, salient shape features of various types are extracted and used to represent curvilinear objects at different levels of abstraction. The searching process is carried out top-down?first at the top level where only numbers of features of the same type are compared, then at the middle level where the geometric constraints among the features are checked, and finally at the bottom level where the parts between the features are considered. The searching space is reduced at each level and finally the most extensive matching operation needs to be applied to only a restricted set of candidates, thus achieving high efficiency. The general scheme has been implemented in two different applications, road image matching and cursive handwriting recognition. Experimental results from both applications are reported. Guidelines for feature selection are also provided to facilitate adaptation of the general scheme to other applications.

Structural Indexing for Character Recognition

In this paper we present a structural method to speed up the character recognition process by reducing the number of the prototypes used during the classification of a given sample. It adopts simplified descriptions of the character shapes and uses a rough classification scheme in order to select the prototypes that most likely will match a given sample. The descriptions are stored in a multilevel data structure adopted to represent the character shape. The lowest level of such a data structure contains the detailed description of the character in terms of its skeletal features. The intermediate one consists of a list of groups of features, each one representing a character component. The upper level, eventually, is an index vector, whose dimension equals the different types of superfeatures. By using this index vector a fast and reliable selection of the prototypes to be considered as candidates for the matching can be obtained. Once this subset has been obtained, the more detailed description based on the skeletal features is resorted and the main classifier activated. Experiments have proved that the method is efficient and correct, since it allows us to select a small subset of prototypes which always contains the right one, thus reducing the classification time without affecting the accuracy of the system.

A line sweep thinning algorithm

We propose a new thinning algorithm based on line sweep procedures. A line sweep is a process where the plane is divided into parallel slabs by lines passing through certain events and then items are processed according to an order of the slabs. Assuming that the contours of the object that are to be thinned have been approximated by polygons, the events are then the vertices of the polygons and the line sweep algorithm looks for pairs of polygon sides that lie within each slab. Since the procedure is applied in both horizontal and vertical direction, possible conflicts may exist among the pairs of polygon sides. A subsequent effort is to resolve the conflicts according to a few generic types into which they can be classified. After the conflict resolution, the object can be decomposed into the regions that can be represented by the skeletons computed from the pairs of polygon sides, and the regions that are the singular parts of the object. Both types of regions can be used as features for subsequent pattern recognition operations.

Line sweep thinning algorithm for feature analysis

In a previous article (Proc. 3rd Int. Conf. Document Anal. and Recogn., Montreal, Canada, pp. 227-30, 1995), we showed that a line sweep algorithm is an efficient means of line thinning. A line sweep is a process that works on polygonal figures and pairs the edges that bound the figure interior from two sides. In this article, we improve and extend this approach in the following way. First, a new method is used for grouping paired edges into regular and intersection regions. The regular regions can be found at the site where pairings between edges exist. Intersection regions, on the other hand, are where such relations cease to exist, due to the fact that pair relations between edges of wide distance are cancelled. Secondly, a salient feature of our new approach is to simultaneously find the set of regular regions that attach to the same intersection region. Such a set is called an intersection set. The output of our algorithm consists of skeletons as well as intersection sets. Both of them can be used as features for subsequent character recognition. Moreover, the line sweep thinning algorithm is efficient in computation as compared with a pixel-based thinning algorithm which outputs skeletons only.

Discrete geometry and Azriel Rosenfeld

Abstract Professor Azriel Rosenfeld did important work in several areas of Image Processing, Pattern Recognition and Computer Vision. I had the good fortune to interact with him on several topics for over 30 years and it is difficult to select a single topic to cover in a short paper. I settled on digital geometry because both our first and last contact dealt with this area. Furthermore Azriel not only did pioneering work on digital geometry, he also continued research on the topic for the rest of his life. Instead of trying to write something on a specific sub-topic of digital geometry, I opted to discuss why the area is so challenging.

A Client-Server Architecture for Document Image Recognition

We propose a client-server architecture for deploying document image recognition applications, especially graphics recognition applications, in large organizations. An example of such an application is presented. We discuss advantages of client-server techniques over the currently available stand-alone tools for document image recognition.