Text Extraction from Document Images- A Review

Ghai, Deepika; Jain, Neelu

doi:10.5120/14559-2661

Cited by 10 publications

(3 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The studies [6], [7], [9], [10], [14] demonstrate achieving high accuracy on text recognition in natural scene and in document images [8], [11], [12]. These are large networks which demands high complexity for training and still fail to detect small sized text (<15 font size).…”

Section: Introductionmentioning

confidence: 99%

Small Text Extraction from Documents and Chart Images

Busa¹,

shahira²,

Lijiya³

2022

Preprint

View full text Add to dashboard Cite

<p>Text recognition is an important area in computer vision which deals with detecting and recognizing text from an image. Optical Character Recognition (OCR) is a saturated area these days and with very good text recognition accuracy. However the same OCR methods when applied to text with small font size like the text data of chart images, the recognition rate is less than 30%. In this work, Our aim is to extract small text in images using the deep learning model, CRNN with CTC loss. The text recognition accuracy is found to improve by applying image enhancement by super resolution prior to CRNN model. We observed that the text recognition rate increases by 24% by applying the proposed method, which involves super resolution and character segmentation using projection profile followed by CRNN with CTC loss. The efficiency of the proposed method shows that further pre-processing on chart image text and other small text images will improve the accuracy further, thereby helping text extraction from chart images.</p>

show abstract

Section: Introductionmentioning

confidence: 99%

Small Text Extraction from Documents and Chart Images

Busa¹,

shahira²,

Lijiya³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…A multiband or raster image consists of line drawings, graphics, pixels, and sketches which are developed by historical document image data, journals scanning, handwritten textual images, printed documental images, printed multi-color books cover and newspaper imagery data. Therefore, there are many challenges in extractions techniques like data can be in low resolution, color bleeding, unknown text color and image with complex background [2].…”

Section: Introductionmentioning

confidence: 99%

Image Classification and Text Extraction using Convolutional Neural Network

Zaman

Bashir

Zaidi

2021

JCBI

View full text Add to dashboard Cite

As luck would have its recent innovations in computer vision grant us to make considerable pace abate the data classification and analysis of enormous data in different organizations like training material, policy guides, and project documents that can be used internally. In addition, cloud service providers are rising in text detection techniques and offer many computer vision contributions included Google Vision, A WS Textract, and Azure OCR. In this paper image classification is achieved by CNN algorithm which is the top choice for image classification. Then using CNN, we extract text data from images. CNN has better performance rate on large datasets as it controls the problem of overfitting. Hence, the accuracy of these algorithms can be enhanced by increasing the epochs and integrating a large dataset. This approach used a methodology to resolve the number of convolution and pooling layers with the number of nodes in network. Lastly, CNN algorithms are better when data is appropriate if it has mortal features that can be analyzed and utilized by algorithm.

show abstract

“…A document image contains various information such as texts, pictures and graphics [5]. These images are obtained by scanning handwritten documents, old documents, printed documents, journals etc.…”

Section: B Document Imagesmentioning

confidence: 99%