Morphological document recovery in HSI space

Abstract

Old documents frequently appear with digitalization errors, uneven background, bleed-through effect etc... Motivated by the challenge to improve printed and handwritten text, we developed a new approach based on morphological color operators using HSI color space. Our approach is composed of a morphological background estimation for foreground/background separation and text segmentation, a background smoothing and a color text recovery. Experimental results carried onto ancient documents have proven that ISH lexicographic order is the most effective to estimate the background and recover ancient texts in uneven and foxed background images. © 2013 Springer International Publishing.