Multi-scale Document Description Using Rectangular Granulometries

Authors:
Andrew D. Bagdanov;Marcel Worring
Affiliations:
-;-
Venue:
DAS '02 Proceedings of the 5th International Workshop on Document Analysis Systems V
Year:
2002

Citing 5
Cited 3

Pattern Spectrum and Multiscale Shape Representation

IEEE Transactions on Pattern Analysis and Machine Intelligence
Model-based morphology: the opening spectrum

Graphical Models and Image Processing
The indexing and retrieval of document images: a survey

Computer Vision and Image Understanding - Special issue on document image understanding and retrieval
Page segmentation using the description of the background

Computer Vision and Image Understanding - Special issue on document image understanding and retrieval
Granulometries and opening trees

Fundamenta Informaticae - Special issue on mathematical morphology

Finite State Machine-Based Optimization of Data Parallel Regular Domain Problems Applied in Low-Level Image Processing

IEEE Transactions on Parallel and Distributed Systems
Detecting cartoons: a case study in automatic video-genre classification

ICME '03 Proceedings of the 2003 International Conference on Multimedia and Expo - Volume 2
Cyclic mathematical morphology in polar-logarithmic representation

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

When comparing documents images based on visual similarity it is difficult to determine the correct scale and features for document representation. We report on new form of multivariate granulometries based on rectangles of varying size and aspect ratio. These rectangular granulometries are used to probe the layout structure of document images, and the rectangular size distributions derived from them are used as descriptors for document images. Feature selection is used to reduce the dimensionality and redundancy of the size distributions, while preserving the essence of the visual appearance of a document. Experimental results indicate that rectangular size distributions are an effective way to characterize visual similarity of document images and provide insightful interpretation of classification and retrieval results in the original image space rather than the abstract feature space.