A structural representation for understanding line-drawing images

Abstract

In this paper, we are concerned with the problem of finding a good and homogeneous representation to encode line drawing documents (which may be handwritten). We propose a method in which the problems induced by a first step skeletonization have been avoided. First, we achieve a vectorization of the image that enables a fine description of the drawing using only vectors and quadrilateral primitives. A structural graph is built based on these primitives extracted from the initial line drawing image. The objective is to manage attributes relative to elementary objects so as to provide a description of the spatial relationships (inclusion, junction, intersection, etc.) that exist between the graphics in the images. This is done by this representation that provides a global vision of the drawings. The capacity of the representation to evolve and to carry high semantical information is highlighted too. Finally, we show how an architecture using this structural representation and a mechanism of perceptive cycles could enable a good quality interpretation of line drawings. key words : vectorization, feature extraction, structural representation, line drawings JY Ramel, N. Vincent, H. Emptoz: A structural representation for the understanding of line drawings

Local Structural Analysis: A Primer

“…This comparison study is only based on a set of significant experiments performed in our laboratories. The skeletonisation based methods are invariant [54], but they only permit the linear shape analysis [45], and are noise sensitive (especially for the junction zone analysis [54]). …”

“…So, this construction is not only a graph factorisation step. After this construction, new structural representations can be created, describing the structural relations between the high-level objects [45]. The Figure 3 (a) gives a use-case for circle reconstruction from skeletonisation based process [23].…”

“…[17] corrects its run graphs with fusion/suppression processings of segmented/isolated runs. [45] corrects its high-level object graphs with a merging processing. …”

A graph-based method for fitting planar B-spline curves with intersections

“…Since medial axis is sensitive to noise, a lot of works study robust generation of clean medial axis. There are also skeleton extraction methods making use of Delaunay triangulations of some sampling points of the shape [23].…”

A Mini Arc Recognition Algorithm Based on the Topological Inference

“…With respect to building the topological relations of the feature raster graphics, it needs some vectorization methods to represent the feature raster graphics. The papers [3,4] present a method based on the quadrilateral to build the relation of the raster graphics. This method is not only ineffective, but also is difficult to represent the complicated graphics.…”