| 摘要 |
An image processor receives color signals representing a color or black-and-white image, typically containing text and non-text areas. A sliding window or swath of the image is processed which progressively moves over the virtual image.A spatial filter is applied to sharpen the image and it is then classified into text and non-text regions. The data from the text regions is subjected to a black text enhancement process in which the color signal from a single channel (here the green channel) is thresholded against two thresholds, T1, T2. The lower (darker) threshold T2 identifies pixels for being set to black, whereas the threshold T1 identifies "support" pixels used in component connectivity analysis. Having defined a connected component using both T1 and T2 pixels, the color statistics of the pixels making up the component are analyzed to determine whether the component should be rendered black. If so, the image data is enhanced by snapping the T2 pixels to black, and snapping a halo of pixels around the black text component to white.
|
