Writer Identification and Writer Retrieval Using the Fisher Vector on Visual Vocabularies

Abstract

In this paper a method for writer identification and writer retrieval is presented. Writer identification is the task of identifying the writer of a document out of a database of known writers. In contrast to identification, writer retrieval is the task of finding documents in a database according to the similarity of handwritings. The approach presented in this paper uses local features for this task. First a vocabulary is calculated by clustering features using a Gaussian Mixture Model and applying the Fisher kernel. For each document image the features are calculated and the Fisher Vector is generated using the vocabulary. The distance of this vector is then used as similarity measurement for the handwriting and can be used for writer identification and writer retrieval. The proposed method is evaluated on two datasets, namely the ICDAR 2011 Writer Identification Contest dataset which consists of 208 documents from 26 writers, and the CVL Database which contains 1539 documents from 309 writers. Experiments show that the proposed methods performs slightly better than previously presented writer identification approaches.

A comprehensive survey of handwritten document benchmarks: structure, usage and evaluation

“…The ground truth data is available in XML format which includes transcription of text, the bounding box of each word and the identity of writer. The database has been used for writer recognition and retrieval [114] and can also be employed for other recognition tasks. A sample image from the database is shown in Fig.…”

Text-Independent Writer Identification Using Improved Structural Features

“…And the bag of word model is used to compute a histogram of MSD (MSDH) as a feature vector.The experimental results demonstrate MSDH gets better performance than the previous SIFT descriptor based feature [16].…”

“…In general, the existing approaches can beclassified into two categories: texture-based approaches and structure-basedapproaches.The texture-based approaches [5][6][7][8]need an amount of handwriting to extract stable and powerful features. Actually,most of time it's unrealistic to collect a large amount of handwriting data.To overcome this problem, more and more structure-based approacheshave been proposed.The structure-based features are more intuitionistic and strong than the texture-based ones and can be roughly divided into contour-based directional features [9,10], local contour pattern based features [11,12], connectedcomponentsbased features [10,[12][13][14], and local descriptors based features [15,16]. This paper analyzes the shortcomings of the contour-based directional features and the local descriptors based features and presents variants of them to improve the performance of writer identification.…”

“…For the local descriptors based features, the SIFT based feature [15,16] is widely used due to the robustness and invariance of SIFT.Since the orientation information of SIFT keypoints is discarded, the SIFT descriptors based feature cannot reflect the orientation characteristic of handwriting which has been proved as an important clue for writer identification as shown in [10].In this work, the orientation information is integrated into the SIFT descriptors by modifying the original SIFT algorithm, called the modified SIFT descriptor (MSD). And the bag of word model is used to compute a histogram of MSD (MSDH) as a feature vector.The experimental results demonstrate MSDH gets better performance than the previous SIFT descriptor based feature [16].…”

Writer Identification for Historical Arabic Documents

“…In our work, we focus on macro features which capture writer characteristics globally like contour-based [2] and textural features [3]. Recently, local feature based approaches which rely on key point descriptors were proposed which showed promising results [4], [5]. A detailed overview about this topic is given in [6].…”