Salvatore Tabbone

A new shape descriptor defined on the radon transform

This paper presents a novel approach to identify complex shapes based on the Radon transform. We propose an adaptation of Radon transform called R-transform, which is invariant to common geometrical transformations. Moreover, to improve the uniqueness of the approach, a binary shape is projected into the Radon space for different levels of the Chamfer distance transform. The accuracy and the efficiency of the proposed algorithm in the presence of a variety of transformations are evaluated within a shape recognition process.

Stable and Robust Vectorization: How to Make the Right Choices

As a complement to quantitative evaluation methods for raster-to-graphics conversion, we discuss in this paper some qualitative elements which should be taken into account when choosing the different steps of ones vectorization method. We stress the importance of having robust methods and stable implementations, and we base ourselves extensivelyon our own implementations and tests, concentrating on methods designed to have few, if any, parameters.

Matching of graphical symbols in line-drawing images using angular signature information

Abstract.In this paper, a method for matching complex objects in line-drawings is presented. Our approach is based on the notion of

Technical symbols recognition using the two-dimensional Radon transform

\mathcal{F}

A system to detect rooms in architectural floor plan images

-signatures, which are a special kind of histogram of forces [17,19,28]. Such histograms have low time complexity and describe signatures that are invariant to fundamental geometrical transformations such as scaling, translation, symmetry, and rotation. This article presents a new application of this notion in the field of symbol identification and recognition. To improve the efficiency of matching, we propose using an approximation of the

A multi-scale edge detector

\mathcal{F}

Histogram of radon transform. A useful descriptor for shape retrieval

-signature from Fourier series and the associated matching.

Vectorization in graphics recognition: to thin or not to thin

We introduce a new method to generate feature vectors to be used in symbol recognition. We propose a new exploitation of the Radon transform to generate relevant-features. The Radon transform is essentially a transformation of an image into a transform plane (/spl rho/,/spl theta/) represented by an accumulator in the discrete case. From the accumulator array we extract a signature (R-signature) which provides global information of a binary shape whatever its type and its form. The signature allows to keep fundamental geometrical transformation like scale, translation and rotation.

Attributed Graph Matching using Local Descriptions

In this article, a system to detect rooms in architectural floor plan images is described. We first present a primitive extraction algorithm for line detection. It is based on an original coupling of classical Hough transform with image vectorization in order to perform robust and efficient line detection. We show how the lines that satisfy some graphical arrangements are combined into walls. We also present the way we detect some door hypothesis thanks to the extraction of arcs. Walls and door hypothesis are then used by our room segmentation strategy; it consists in recursively decomposing the image until getting nearly convex regions. The notion of convexity is difficult to quantify, and the selection of separation lines between regions can also be rough. We take advantage of knowledge associated to architectural floor plans in order to obtain mostly rectangular rooms. Qualitative and quantitative evaluations performed on a corpus of real documents show promising results.

Graph Matching Based on Node Signatures

Abstract A multi-scale edge detector with subpixel accuracy is described. A subpixel Laplacian edge detector, recursively implemented, is run at different scales and the recovered edge information is combined. The multi-scale edge detection is based on the behavior of edges in scale space and takes into account their physical phenomena. With this purpose in mind, four-step edge models are considered: the ideal, the blurred, the pulse and the staircase. It is emphasized that the use of two scales (the larger and the smaller) is sufficient for good edge detection. Furthermore, a set of rules is derived for combining edge information obtained from a Laplacian detector which has some special properties. However, this type of edge detector gives at least two classes of false edges, one of which cannot be eliminated by the usual thresholding methods. An appropriate thresholding algorithm is given taking into account the origin of the false edges and their behavior in scale space.