Current Research Activities

Currently, my research interests are focused on the following cues:

Graphics recognition. It is a subfield of document analysis and understanding. It deals with symbols, charts, diagrams, logos, etc. Some outstanding applications of graphics recognition are the matching of fingerprints, the interpretation of flow charts and diagrams, the analysis of musical scores for conversion to MIDI, the conversion of engineering designs to a CAD-compatible representation, the interpretation of architectural drawings, the analysis of maps in GIS environents, the interpretation of electronic circuit diagrams, etc. My current research is concerned about one of the central problems of graphics recognition: symbol recognition. I have stated the problem in terms of a graph matching formulation. The algorithms have been developed for an architectural drawings understanding application (see further information in the HDICAD homepage).

Structural pattern recognition. This is a field started at the end of 1960's and the beginning of the 1970's. Nowadays, it is still an active area of research and a number of applications have been done and are progressing using these techniques. The basis of the structural methods is the relational organization of low-level features into higher-level structures. This relational organization is represented by means of symbolic data structures like strings, trees, graphs or arrays instead of vectors of numbers that are used in the statistical pattern recognition approach. These data structures also allow a hierarchical organization of the information. Thus, two major attributes define the structural pattern recognition field, namely, a hierarchical structure of a pattern in terms of simpler subpatterns, and a set of relations which may exist between different subpatterns. I am currently working on two particular problems:

Error-tolerant (sub)graph isomorphism. Graphs are clearly the most widely used data structure to represent n-dimensional prototypes. Thus, a pattern may be recognised by means of a graph matching procedure. Since the presence of noise and distortion in input images is a usual fact in Computer Vision, an exact graph matching procedure appears to be unreliable. An error model have to underly the matching process. I am developing some inexact subgraph isomorphism methods based on the string edit distance and Region Adjacency Graphs representations.

Inference of structural models. One of the fields where structural pattern recognition has an intensive action is structural texture analysis. A structural texture is characterized by a repetitive element called texel repeated according to a set of placement rules. Graph-based formal grammars are useful structures to represent such repetitive structures. Two problems arise, namely, texture recognition and the inference of a structural model for texture representation.

 

Publications 

1999

J. Lladós, J. López-Krahe, E. Martí. A Hough-based method for hatched pattern detection in maps and diagrams. International Conference on Document Analysis and Recognition (ICDAR'99). Bangalore (India), 20-22 september, 1999. Accepted.

J. Lladós, F. Lumbreras, X. Varona. A multidocument platform for automatic reading of identity cards. VIII National Symposium on Pattern Recognition and Image Analysis. Bilbao, 1999. Accepted.


1998

J. Lladós, G. Sánchez, E. Martí. A string-based method to recognize symbols and structural textures in architectural plans. Graphics Recognition, Algorithms and Systems. K. Tombre and A.K.Chhabra (eds). Lecture Notes in Computer Science, 1389, pp. 91-103. Springer-Verlag, 1998.


1997

J. Lladós, H. Bunke, E. Martí. Finding rotational symmetries by cyclic string matching, Pattern Recognition Letters, 18(14):1435-1442, 1997.

J.Lladós, Combining Graph Matching and Hough Transform for Hand-Drawn Graphical Document Analysis. Application to Architectural Drawings , PhD Dissertation, November, 1997.

J. Lladós, G. Sánchez, E. Martí. A string-based method to recognize symbols and structural textures in architectural plans. GREC'97 (2nd. IAPR Workshop on Graphics Recognition), Nancy (France), August, 1997.

G. Sánchez, J. Lladós, E. Martí. Segmentation and analysis of lineal textures in plans. VII National Symposium on Pattern Recognition and Image Analysis, CVC-UAB, Barcelona, 1997.

J. Lladós, E. Martí. Playing with error-tolerant subgraph isomorphism in line drawings. VII National Symposium on Pattern Recognition and Image Analysis, CVC-UAB, Barcelona, 1997.

J. Lladós, J. López-Krahe, E. Martí, A system to understand hand-drawn floor plans using subgraph isomorphism and Hough transform, Machine Vision and Applications, 10(3):150-158, 1997.

J. Lladós, H. Bunke, E. Martí, Using cyclic string matching to find rotational and reflectional symmetries in shapes. Intelligent Robots - Sensing, modeling and planning. R.C. Bolles, H. Bunke, H. Noltemeier (eds), pp. 164-175, World Scientific, 1997.


1996

J.Lladós, J. López-Krahe, E. Martí, Hand Drawn Document Understanding Using the Straight Line Hough Transform and Graph Matching, 13th. International Conference on Pattern Recognition, Viena (Austria), 1996.

J.Lladós, Interpretació de dibuixos lineals fets a ma alçada mitjancant isomorfisme entre subgrafs i transformació de Hough, CVC Technical Report #010, March 1996.


1995

J.Lladós, Interpretacio de dibuixos lineals mitjancant tecniques d'isomorfisme entre grafs, I Trobada de Joves Investigadors, Bellaterra (Barcelona), 1995.

J.Lladós, E.Martí, Structural recognition of hand drawn floor plans, VI National Simposium on Pattern Recognition and Image Analysis, Cordoba, 1995.


1993

J.Lladós, E.Martí, J.Regincós Interpretacion de diseños a mano alzada como técnica de entrada a un sistema CAD en un ámbito de arquitectura, III National Conference on Computer Graphics, Granada, 1993.

 

Back to Josep Lladós Homepage