Image patches analysis for text block identification

Abstract

In this paper, we propose a novel text block identification method for ancient document understanding. Unlike traditional top-down and bottom-up approaches, our method is based on supervised learning on the patches of document images, which can be considered as an intermediate level method but integrates essential advantages of both the top-down and the bottom-up strategies. In our method, the document images are firstly partitioned into small patches, and then positive and negative patches are selected to form an active training set. Gabor features are extracted on each patch, while multi-linear discriminant analysis (MDA) is employed to reduce the dimensionality of the data. To deal with unseen documents, a random forest classifier is learned on the new representations of the patches. Compared to traditional approaches, our method can not only capture local texture features of each patch, but also preserve the global information of the training images. Furthermore, MDA is guaranteed to learn a low dimensional tensor subspace, which significantly avoids the curse of dimensionality dilemma. Moreover, the random forest classifier can automatically select useful features and deliver satisfactory identification results. Extensive experiments on some scripts of ancient document images demonstrated the effectiveness of our method.

An overview on data representation learning: From traditional feature learning to recent deep learning

“…Except feature learning algorithms mentioned above, there are many other feature learning methods, such as independent component analysis (ICA) [46], canonicalcorrelation analysis (CCA) [47], ensemble learning based feature extraction [48], multitask feature learning [49], and so on. Moreover, to directly process tensor data, many tensor representation learning algorithms have been proposed [50,51,52,23,53,54,55,56,57,58]. For example, Yang et al proposed the 2DPCA algorithm and shew its advantage over PCA on face recognition problems [50], while Ye, Janardan and Li proposed the 2DLDA algorithm, which extends LDA for two-order tensor representation learning [51].…”

Use of SLIC superpixels for ancient document image enhancement and segmentation

“…Zhong and Cheriet used the dimensionally reduced multi-channel Gabor filters for text block identification on image patches from ancient documents [49]. The authors extracted 28 Gabor filters from image patches in their experiments, where 7 spatial frequencies (…”