Abstract: A television transmission system for digital picture signals in which each picture is split up into sub-pictures of NxN pixels for the purpose of reducing the bit rate. Each sub-picture is subjected to a two-dimensional discrete cosine transform (DCT) for determining basic picture weighting factors (coefficients) which are converted by adaptive coding into code words of variable length. This two-dimensional DCT is realized by multiplying the pixels of each row of the sub-picture by the NxN DCT matrix (one-dimensional transform). Consequently, an NxN array of product elements is obtained. The transposed version thereof is once more multiplied by the NxN DCT matrix. Such a one-dimensional transform includes three main steps. In a first main step, the pixels of a row are converted in a first iteration step, into N/2 sum elements u.sub.1,i and N/2 difference elements v.sub.1,i. In a subsequent iteration step, the N/2 sum elements u.sub.1,i are converted into N/4 further sum elements u.sub.