Multilingual scene character recognition with co-occurrence of histogram of oriented gradients

Tian, Shangxuan; Bhattacharya, Ujjwal; Lu, Shijian; Su, Bolan; Wang, Qingqing; Wei, Xiaohua; Lu, Yue; Tan, Chew Lim

doi:10.1016/j.patcog.2015.07.009

Cited by 134 publications

(50 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From Table 1, we can see that our method is superior to other published methods including HOG based and CNN (using fully-connected layer features) based methods. Compared with HOG+SVM [19] and Co-HoG [20] which encode the spatial information by considering the co-occurrence of orientation pairs, our method achieves superior performance. Compared with SED [9], DSEDR [19], DMSDR [19] and Stoke Bank [21], which utilize the HOG as the local descriptors to capture stroke structure information, the proposed BCA-FV method outperforms them by more than 3% (8%), 2% (4%), 3% (9%) and 5% (10%) on the ICDAR2003 (Chars74K) database, respectively.…”

Section: Comparison With Other Methodsmentioning

confidence: 99%

“…We also conduct experiments on the Chinese scene character database "Pan+ChiPhoto" [20] and adopt the same experimental setup as described in [20].…”

Section: Databases and Implementation Detailsmentioning

confidence: 99%

“…ICDAR Chars74K "Pan+ChiPhoto" HOG+SVM [19] 77.00 62.00 59.20 Co-HoG [20] 80.50 -64.30 Stoke Bank [21] 79.80 65.90 -SED [9] 82.00 67.10 -DSEDR [19] 82.60 71.80 -DMSDR [19] 81.70 66.10 -CNN+softmax [22] 81.57 73.52 53.40 FV+SVM [22] 84.40 74.80 -MCA-FV [23] 83.40 -76.70 CNN [14] 86.80 -61 BCA-FV gains higher accuracy than the CSM-FV, because our method can automatically select the useful descriptors and remove the less importance ones, leading to a more powerful representation. The CNN [14] achieves 86.80% on the ICDAR2003 database for character recognition which partly attributes to the large amount of additional training data, i.e., 107k.…”

Section: Algorithmmentioning

confidence: 99%

See 2 more Smart Citations

Bilateral Convolutional Activations Encoded with Fisher Vectors for Scene Character Recognition

Zhong

Wang

Liu

et al. 2018

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYA rich and robust representation for scene characters plays a significant role in automatically understanding the text in images. In this letter, we focus on the issue of feature representation, and propose a novel encoding method named bilateral convolutional activations encoded with Fisher vectors (BCA-FV) for scene character recognition. Concretely, we first extract convolutional activation descriptors from convolutional maps and then build a bilateral convolutional activation map (BCAM) to capture the relationship between the convolutional activation response and the spatial structure information. Finally, in order to obtain the global feature representation, the BCAM is injected into FV to encode convolutional activation descriptors. Hence, the BCA-FV can effectively integrate the prominent features and spatial structure information for character representation. We verify our method on two widely used databases (ICDAR2003 and Chars74K), and the experimental results demonstrate that our method achieves better results than the state-of-the-art methods. In addition, we further validate the proposed BCA-FV on the "Pan+ChiPhoto" database for Chinese scene character recognition, and the experimental results show the good generalization ability of the proposed BCA-FV.

show abstract

Section: Comparison With Other Methodsmentioning

confidence: 99%

“…We also conduct experiments on the Chinese scene character database "Pan+ChiPhoto" [20] and adopt the same experimental setup as described in [20].…”

Section: Databases and Implementation Detailsmentioning

confidence: 99%

Section: Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Bilateral Convolutional Activations Encoded with Fisher Vectors for Scene Character Recognition

Zhong

Wang

Liu

et al. 2018

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…Chinese character dataset : Pan+ChiPhoto dataset [25]. It is built by the combination of two datasets: ChiPhoto and Pan Chinese Character dataset.…”

Section: Datasetsmentioning

confidence: 99%

Boosting scene character recognition by learning canonical forms of glyphs

Wang

Lian

Tang

et al. 2019

IJDAR

View full text Add to dashboard Cite

As one of the fundamental problems in document analysis, scene character recognition has attracted considerable interests in recent years. But the problem is still considered to be extremely challenging due to many uncontrollable factors including glyph transformation, blur, noisy background, uneven illumination, etc. In this paper, we propose a novel methodology for boosting scene character recognition by learning canonical forms of glyphs, based on the fact that characters appearing in scene images are all derived from their corresponding canonical forms. Our key observation is that more discriminative features can be learned by solving specially-designed generative tasks compared to traditional classification-based feature learning frameworks. Specifically, we design a GAN-based model to make the learned deep feature of a given scene character be capable of reconstructing corresponding glyphs in a number of standard font styles. In this manner, we obtain deep features for scene characters that are more discriminative in recognition and less sensitive against the above-mentioned factors. Our experiments conducted on several publicly-available databases demonstrate the superiority of our method compared to the state of the art.

show abstract

“…For each pixel in an image block, the gradient orientations of the pixel pair formed by its neighbour and itself are examined. The Co-HOG descriptor is broadly used in object detection because of precisely representing the important uniqueness of an object structure, therefore CO-HOG descriptor is more applicable for real time applications such as detection of pedestrian [12], detection of human body parts [4], detection of objects in natural scene images [13], recognition of text in natural scene [15], recognition of characters in natural scenes [16], and recognition of characters in multilingual scene images [17].In this paper, we propose a novel segmentation based word spotting method for handwritten document images using Co-HOG descriptor. Initially, we divide a word image into number of blocks, then, Co-HOG features are extracted from each block.…”

mentioning

confidence: 99%

Segmentation Based Word Spotting Method for Handwritten Documents

Prabhakar

2017

IJARCSSE

View full text Add to dashboard Cite

Abstract-In this paper, we present a segmentation based word spotting method for handwritten document images using Co-occurrence Histograms of Oriented Gradients (Co-HOG) descriptor. The drawback of Histogram of Oriented Gradients (HOG) is that HOG ignores spatial information of adjacent pixels where as the Co-HOG take into account spatial contextual information by capturing the co-occurrence of orientation pairs of neighbouring pixels. In order to construct Co-HOG descriptor for word spotting, we divide a word image into blocks and Co-HOG features are extracted from each block and finally concatenate them to form a feature descriptor. The proposed method is evaluated using precision and recall metrics through experimentation conducted on popular GW dataset and confirmed that our method outperforms for this dataset.Keywords-Word spotting, Character Recognition, George Washington, Dynamic Time Warping, Hidden Markov Models I. INTRODUCTIONRecently Document Image Analysis is become one of dynamic research field which draws an attention of researcher due to its complexity and growing requirement for accessing the content of digitized information. Optical Character Recognition (OCR) has been explored for a few decades with massive accomplishment which facilitates to automate human procedure. OCR techniques usually recognize words by processing fonts independently and works well with machine printed fonts against clean environment. Generally, big quantity of document images are accumulated in digital libraries, and processing of these documents with the help of OCR requires high computation rate due to difficulty involved in understanding the page layout of digitized documents, irregular writing manner, dull ink, stained paper and other adverse factors. In order to overcome these problems, researchers have proposed a method called word spotting. Word spotting method is a moderately new alternative for text recognition and retrieval in digitized printed and handwritten documents.Handwritten word spotting is the pattern classification mission which consists of detecting given query word in handwritten document images. The word spotting in handwritten documents is not completely solved due to various challenges posed by handwritten documents. Hence, we focused on word spotting in handwritten documents rather than printed documents. Generally, a word spotting method consists of three main modules: pre-processing, feature extraction and feature matching. Among them, feature extraction is one of most important factors for achieving high retrieval performance, because of feature with strong discriminative information can be well classified even using with simplest classifier.The literature investigation exposes that HOG descriptor is extensively used in numerous recognition applications because of its discriminative capability compared to other existing feature descriptors. The HOG descriptor is developed by Dalal Importantly, HOG considers orientation of only isolated pixels, whereas spatial information of adjacent pixels i...

show abstract

Multilingual scene character recognition with co-occurrence of histogram of oriented gradients

Cited by 134 publications

References 34 publications

Bilateral Convolutional Activations Encoded with Fisher Vectors for Scene Character Recognition

Bilateral Convolutional Activations Encoded with Fisher Vectors for Scene Character Recognition

Boosting scene character recognition by learning canonical forms of glyphs

Segmentation Based Word Spotting Method for Handwritten Documents

Contact Info

Product

Resources

About