Stamps Detection and Classification Using Simple Features Ensemble

Abstract

The paper addresses a problem of detection and classification of rubber stamp instances in scanned documents. A variety of methods from the field of image processing, pattern recognition, and some heuristic are utilized. Presented method works on typical stamps of different colors and shapes. For color images, color space transformation is applied in order to find potential color stamps. Monochrome stamps are detected through shape specific algorithms. Following feature extraction stage, identified candidates are subjected to classification task using a set of shape descriptors. Selected elementary properties form an ensemble of features which is rotation, scale, and translation invariant; hence this approach is document size and orientation independent. We perform two-tier classification in order to discriminate between stamps and no-stamps and then classify stamps in terms of their shape. The experiments carried out on a considerable set of real documents gathered from the Internet showed high potential of the proposed method.

Two-stage approach to extracting visual objects from paper documents

“…It applies Hough line and circle transforms, color segmentation and heuristic techniques. As it was stated in above-mentioned literature survey, logo detection is a very similar problem and can be solved with a little tweak to our previously presented solution [5]. Other authors propose to use key-point analyzing algorithms like ScaleInvariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF) and Features from Accelerated Segment Test (FAST) or Angular Radial Transform (ART).…”

“…It is based on classification of characteristic features, often in a scheme "one versus all". In our previous works [5,9] a similar problem of stamp detection and recognition was described in detail. It applies Hough line and circle transforms, color segmentation and heuristic techniques.…”

“…So-called low-level features are a result of our previous research on stamp detection and recognition [5,9]. This approach shares common features with measures proposed by Haralic et al Created feature vector contains eleven values, namely contrast, correlation, energy and homogeneity calculated in the same way as in case of GLCM matrix.…”

“…It could be an integral part of any content-based image retrieval system, or simply a filter that would select only documents containing specific elements [4], segregate them in terms of importance (colored documents containing stamps and signatures are more valuable than monochromatic ones, which suggest a copy [5,10]), etc. Presented approach is document-type independent; hence, it can be applied to any formal documents, diploma, newspapers, postcards, envelopes, bank checks, etc.…”

Machine learning methods analysis in the document classification problem

“…Наступний підхід до виявлення печаток у документах, представлений у [2], використовує поєднання деяких простих характеристик зображення. Алгоритми машинного навчання (такі як метод kнайближчих сусідів, метод опорних векторів, випадкові ліси), що використовуються для виявлення печаток, обробляють інформацію про зображення, в якому закодували початкову модель RGB у модель, що представляє зображення як поєднання Y, Cb, Cr, де кожна з компонент є сумою значень RGB, перемножених на сталі коефіцієнти, після чого зображення бінаризується.…”