First, image data is converted using a first table storing conversion values in addresses corresponding to at least input values: u(k) represented by u(k)=ab.sup.-k (where a and b are constants and k is 0, 1, 2, . . . m) of all input values and linear interpolation. Next, a second table storing conversion values set in response to the image data provided by conversion using the first table in addresses corresponding to at least input values: v(k) represented by v(k)=ck+d (where c and d constants and k is 0, 1, 2, . . . n) of all input values is stored in memory. Next, the image data provided by conversion using the first table is again converted using the second table and linear interpolation.