Methods to Extract Character Regions from Oracle Bone Inscriptions

Abstract

Oracle bone inscriptions are the character used in the ancient Chinese dynasty more than 3000 years old. Deciphering OBIs is very important for study of ancient China and Kanji history and so on. Recently, various methods of OBI recognition by computer are proposed, but they assume a region of oracle bone image including exactly one OBI character is given. We investigate extraction of character regions from an oracle bone image and has proposed a method using general noisereduction filters and a Gabor filter so far. In this manuscript, the effect of the median filter and the local binarization filter is evaluated in addition to our previous method. Using Gaussian and oval Gaussian as region extraction filters in addition to Gabor filter, false reject rate (FRR) and false acceptance rate (FAR) of character regions are measured for each noise-reduction filters. Noise reduction using local binarization achieves about 10% lower FAR compared to previous one.