Extraction of Character String Areas from Color Scenery Image Using Extended Psychological Potential Field

As digital cameras have been used widely, a technology to extract features of character string areas from scenery (nature) image is required. However, it is very difficult to extract character string areas from a complex background of scenery image as compared with a relatively simple background of documentary image. Human can extract character string areas from scenery image having complex background. Therefore, a fairly high degree of accuracy in the extraction of character string areas may be attained by imitating the human visual processing system. In this paper we proposed the method to extract character string areas from color scenery images using the extended psychological potential field (imitating the human visual processing system). The conventional psychological potential field method has the problem that it is hard to be applied to multilevel images such as color scenery images. This paper proposes a new extended psychological potential field method that has improved the conventional method so that it can be applied to multilevel images such as color scenery images. The new extended psychological potential field method proved that processing results closer to the human visual processing system could be obtained by adopting the shielding effects in accordance with the pixel value and distance. 1 Flow of Proposed Method nary images in which the brightness difference is constant. However, color scenery images discussed in this paper need to determine H because the brightness differences between figures and the background are not constant. This paper indicates figure clearness H using the color difference (differences between colors) between figures and the background. The International Commission on Illumination (CIE) recommends the u n i f o y .cqlor space and color difference formula by CIE LAB (L a b ) as methods for quantitatively indicating the c0,Io.r flifference between two colors. This paper uses the L a b color system as color space and defines figure clearness H a s shown in formula 1 (Fig. I).