Detection and rendering of text in tinted areas

摘要: A method and apparatus improves digital reproduction of a compound document image containing half-tone tint regions text and/or graphics embedded within the regions. The entails determining local average pixel value for each in image, then discriminating classifying based on values, text/graphics pixels from pixels. Discrimination can be effected by calculating range averages neighborhood surrounding pixel; edge gradients values; or approximating second derivatives values averages. Text/graphics are rendered using rendering appropriate that type pixel.