HISTOGRAM E QUALIZATION SMOOTHING FOR DETERMINING THRESHOLD ACCURACY ON ANCIENT DOCUMENT IMAGE BINARIZATION

Abstract

Ancient documents are inheritance that must be preserved. The
documents contain historical, scientific, social, religious information, etc.
Converting ancient documents into digital image formats is one of ways to
preserve the inheritance and can be stored into a computer. However,
images of ancientdocuments have many blemishes caused by age,
moisture, flood, etc. Therefore, special techniques are needed for those
images to be restored and can improve the legibility of the ancient
documents’ images. In this study, the image restoration process uses
separation of background and foreground/text on histogram equalization
such as research conducted by Fitri Arnia in 2008. Through histogram
equalizationimages can be seen the distribution of pixels from the intensity
of black color "0" to white "1". The distribution of pixels on histogram
equalization describes the curves of foreground/text and curves of
background. Among the histogram curves, the determination of
thresholdvalues can be done so as to clarify the foreground/text and
background areas on images of ancient documents. The lowest point
between the two curves is the lowest pixel (local minima) which is used as
the threshold value. However, the selection of such threshold values in
some cases is very difficult to determine because there are still many
fluctuations in the curve at the lowest curve. Therefore, this study
proposesa histogram smoothing method in the ancient documents’ images
to minimize curvature fluctuations and to determine more accurate
threshold values. In this research, average filtering method is used for
smoothing the histogram image. This filter successfully refines the
histogram and makes the image of the restoration or binary image display
the value of the ancient document image readability increases.
Keywords: HistogramEqualization, Smoothing Histogram, Average
Filtering, Thresholding