Abstract: Interframe compression of moving image data is achieved by a combination of vector coding and scalar coding of transform coefficient differences formed by the subtraction of previous image transform coefficients from corresponding present image transform coefficients along a Zig-Zag path in the arrays of coefficients. More efficient scalar codes are used instead of vector codes for transmission of coefficient data where the previous image coefficients are non-zero along the path. Vector codes are used where a non-zero value in the previous image transform array has not been established but a non-zero value coefficient occurs in the present image transform array. Vector coding efficiency is improved by modified run length coding.