Word Segmentation for Document Images by Successively Merging Adjacent Character Bounding Boxes by Iterative Dilation

A new method of word segmentation for document images is presented. The method uses the bounding box regions to enclose the letters (characters) of the words and then the resulting letter spaces are progressively filled to merge the character bounding boxes to get the word bounding boxes. The method holds good for inclined and irregularly distributed words. The proposed method completely avoids the line segmentation process which normally precedes word segmentation in traditional methods. Keywords— Bounding boxes, Connected components, Horizontal Dilation, Character spacing, Word bounding boxes, Word segmentation, Word spacing.