Proposed Framework For Handwritten Historical Document Essays
Proposed framework for handwritten historical document binarization is show in figure 2. Proposed framework consist color to gray conversion, contrast measurement, pixel segmentation based back ground elimination and refine fore- ground image by filter out rough foreground image.
Grayscale Conversion: - Proposed framework initially applies grayscale conversion [b8] for eliminating noisy areas, smoothing of background texture as well as contrast enhancement between background and text areas as show in figure.
Fig 1: Input handwritten image
Contrast Measure: Subsequently proposed frame apply Markov random field for contrast measure of a pixel against its background. And statically calculate threshold that’s lead to segment out pixel into three category document background pixels, document foreground (text) pixels, and uncertain pixels. Foreground pixel (darker) usually has a positive value of contrast measure , and most background pixels (lighter) have negative, zero or small positive values. A larger window size yields a better detection of thicker lines patterns while it is difficult to separate very close lines.
Intuitively, this tends to separate the ink from the bottom due to the divergence of the gradient of Markov measurement. Therefore, it will be positive intensity valleys (ink) and negative current peaks or plateaus (bottom). The image contrast is evaluated by the following equation