14.01.2014 Views

Segmentation of heterogeneous document images : an ... - Tel

Segmentation of heterogeneous document images : an ... - Tel

Segmentation of heterogeneous document images : an ... - Tel

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

tel-00912566, version 1 - 2 Dec 2013<br />

Figure 1.2: Left; A <strong>document</strong> page that is segmented incorrectly by ABBYY<br />

FineReader 2011 <strong>an</strong>d Right; The same <strong>document</strong> segmented by our method.<br />

page. The proposed method is based on both intrinsic <strong>an</strong>d extrinsic features<br />

that have the adv<strong>an</strong>tages <strong>of</strong> both connected component based <strong>an</strong>d<br />

block-based methods in detecting graphical elements inside a <strong>document</strong><br />

image.<br />

2. Identification <strong>of</strong> side notes in historical <strong>document</strong> <strong>images</strong> as a new step in<br />

<strong>document</strong> image <strong>an</strong>alysis.<br />

Although due to the adv<strong>an</strong>ces in page segmentation algorithms during<br />

the last decade, page segmentation methods c<strong>an</strong> segment multi-column<br />

<strong>document</strong>s correctly, but problems still exist in some area such as detecting<br />

side notes when they are situated close to the main text. Figure 1.3 shows<br />

the result <strong>of</strong> page segmentation with ABBYY FineReader 2011 that has<br />

gone wrong. Note that when the segmentation is wrong, the result <strong>of</strong><br />

OCR is also not impressive. Figure 1.4 demonstrates the result <strong>of</strong> page<br />

segmentation, obtained from our method in the form <strong>of</strong> paragraphs for<br />

the same image.<br />

3. A powerful framework for text region detection <strong>an</strong>d column separation<br />

that provides the ability <strong>of</strong> inducing prior knowledge about the <strong>document</strong><br />

into the detector.<br />

4. A new trainable method for grouping text lines into paragraphs based on<br />

a binary tree model that maximize the probability <strong>of</strong> preserving groups<br />

6

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!