Spotting a Target Advertisement in a Source Document Image

Many times one would search for a particular content in an entire document image, such as if a company has released an advertisement, the concerned manager would be interested in locating the advertisement in a newspaper. This problem is termed as advertiser 's problem. The target text component is a-priori known in terms of its structure and content, however the complete details of the layout where the target image appears in a source document image will not be known. Given the fact that font size and font style employed by a source document is generally known before hand, the problem is to locate the target image in the source document image. The major objective of this paper is to accomplish this task of spotting the target text component with the additional and the major constraint that for such a purpose any language processor such as OCR neither should be employed nor would be available. For this purpose the components in a document image are characterized in terms of energy values called entropy [4, 9]. The matching of the entropy is proposed for locating the target component in a source document image.