Segmentation of uniform-coloured text from colour graphics background

An approach for separating uniform-coloured text strings (i.e. the characters in a text string are of the same colour) from colour graphic background images is presented. This approach overcomes various difficulties that arise due to colour complexity of graphic backgrounds and disturbance from background graphic objects. With this approach, the character-like components in a created binary edge image are first located and then divided into subgroups according to their colours and locations. Finally, a set of discriminating rules is used to identify the `real´ text strings among the resulting groups of character-like components. Experimental results demonstrated that this approach can successfully separate Chinese and English text strings of various colours from a variety of colour graphic backgrounds