Improvement on separating halftones from text information in size reduced document

Separating halftones from the text information is important for regeneration of a size reduced copying or facsimile document. In the present study, an effective algorithm has been presented for extracting halftones from a longitudinally reduced greyscale image. The approach described in the present paper not only refers to the size expansion-reduction procedures, but also includes the edge emphases and smoothing processes for segmenting a mixed text/halftone document. It has been demonstrated that most of the halftones in a size reduced image can be distinguished from the text region while the misdetections of small letters can be effectively restrained by use of the proposed methodology. The procedure is also thought to be useful for the colour image segmentation.