Existed historical pictures contain more or less granular noise and shading on the background, and that they show blurred contours. The blurred picture would adversely affect to the image processing by computer. A method is proposed in this paper to improve adaptive visibility for blurred image of ancient characters written on the "Paper Squeezes". Images are calculated to extract the character edge, using both gradient operator and wavelet transform. Two images are superposed to emphasize the edge pattern. An image space is divided into small square blocks as sub-images. These are compared with template blocks using image correlation, and are classified into eight patterns. Desirable results in terms of the experiment were obtained for the adaptive visibility improvement.