Search

This paper examines the uses of text and its representation on Web documents in terms of the challenges in its interpretation. Particular attention is paid to the significant problem of non-uniform representation of text. This non-uniformity is mainly due to the presence of semantically important text in image form as opposed to the standard encoded text. The issues surrounding text representation in Web documents are discussed in the context of colour perception and spatial representation. The characteristics of the representation of text in image form are examined and research towards interpreting these images of text is briefly described.

This paper describes a complete approach to the segmentation and extraction of text from Web images for subsequent recognition, to ultimately achieve both effective indexing and presentation by non-visual means (e.g., audio). The method described here (the first in the authors’ systematic approach to exploit human colour perception) enables the extraction of text in complex situations such as in the presence of varying colour (characters and background). More precisely, in addition to using structural features, the segmentation follows a split-and-merge strategy based on the Hue-Lightness- Saturation (HLS) representation of colour as a first approximation of an anthropocentric expression of the differences in chromaticity and lightness. Character-like components are then extracted as forming textlines in a number of orientations and along curves.

There is a significant need to objectively evaluate layout analysis (page segmentation and region classification) methods. This paper describes the Page Segmentation Competition (modus operandi, dataset and evaluation criteria) held in the context of ICDAR2003 and presents the results of the evaluation of the candidate methods. The main objective of the competition was to evaluate such methods using scanned documents from commonly-occurring publications. The results indicate that although methods seem to be maturing, there is still a considerable need to develop robust methods that deal with everyday documents.

There is a significant need to extract and recognise the semantically-important text contained in images on Web pages. This paper proposes a new approach to text extraction from this special class of images. The method attempts to emulate closer than before the way humans perceive colour differences in order to differentiate between text and background regions. Pixels of similar colour (as humans see it) are merged into components and a fuzzy inference mechanism (using connectivity and colour distance features) is devised to group components into larger character-like regions.

Indexing and searching for WWW pages is relying on analysing text. Current technology cannot process the text embedded in images on WWW pages. This paper argues that this is a significant problem as text in image form is usually semantically important (e.g. headers, titles). The results of a recent study are presented to show that the majority (76%) of words embedded in images do not appear elsewhere in the main text and that the majority (56%) of ALT tag descriptions of images are incorrect or do not exist at all. Research under way to devise tools to extract text from images based on the way humans perceive colour differences is outlined and results are presented.

There is a significant need to analyse the text in images on WWW pages, both for effective indexing and for presentation by non-visual means (e.g., audio). This paper argues that the extraction of text from such images benefits from an anthropocentric approach in the distinction between colour regions. The novelty of the idea is the use of a human perspective of colour perception in preference to RGB colour space analysis. This enables the extraction of text in complex situations such as in the presence of varying colour and texture (characters and background). More precisely, characters are extracted as distinct regions with separate chromaticity and/or luminance by performing a layer decomposition of the image. The method described here is the first in our systematic approach to approximate the human colour perception characteristics for the identification of character regions. In this instance, the image is decomposed by performing histogram analysis of Hue and Luminance and merging in the HLS colour space.