Image signal filtering

Abstract

An image signal is filtered in one or two-dimensions for obtaining an unsharp signal in a detail emphasis process of an image reproduction system. A one-dimensional image signal filtering is carried out in one of the main and sub-scanning directions so that the first K image signals and the other K image signals out of 2K image signals are respectively summed to obtain first and second sum signals, which are subjected to a cumulus computation to obtain a filtered signal U.sub.2 expressed by: ##EQU1## for (2K-1) image signals d.sub.i (i: K, K-1, . . . 1, 0, -1, -k+2 In this regard, a cumulus computation refers to computation expressed by p.sub.n -Q.sub.n +R.sub.n-1 =R.sub.n, wherein P.sub.n is an addition signal and Q is a subtraction signal where n is the input order of each of the signals and R.sub.n-1 is the result of the previous computation. By filtering the thus-obtained (2K-1) one-dimensionally filtered signal U.sub.2i (i=K, . . . 1,0, -1, scanning direction factors in the same manner, a two-dimensionally filtered signal W.sub.2, expressed by: ##EQU2## is obtained to be used as an unsharp signal for the abhove-mentioned detail emphasis process. In this connection, the number 2K-1 corresponds to a mask size, which can be arbitrarily determined by adjusting the value K. By carrying out the cumulus computation in a plurality of stages, a weight coefficient of multidimensional function with respect to the value i can be imposed on image signals.