It is well known that the normalization of a pattern is very important in pattern recognition. There are some papers which have been published about a method of the normalization, but no powerful method for this procedure. For this cause, a explanation is given to a improved and powerful method for the normalization by the use of projective geomerty. The basic conception of this method have been proposed by G. Nagy, et. al. in 1970 and applied to a quantized pattern without thinning procedure. In our method, this is applied to thinned pattern and normalized. Moreover, the normalized pattern is lowered in dimension and reduced in size. As the result of this procedure, we can conjecture that the obtained pattern is better than earlier methods and that the memory capacity is saved in computer oriented pattern recognition. Computer-simulated experiments are carried into effect in order to test the usefulness of the present method and some results are shown.