Abstract: A method and apparatus is described which provides improved error correction for reading a PDF417 or other machine-readable symbol having a data region with symbol characters representing encoded data. A processor receives the encoded data read from the PDF417 symbol and executes an error correction routine. The error correction routine includes a two dimensional discrete Fourier transform with a reduced number of arithmetic operations. The transform is a variant of the Good-Thomas FFT performed over the Galois field GF(929), and includes a set of first index values i.sub.1 (consisting of all integers from 0 to 28 inclusive) and a second set of index values i.sub.2 (consisting of all integers from 0 to 31 inclusive). For a given first index value, the transform is performed only over a subset of the second index values. The subset of second index values is the first k values of a sequence given by i.sub.1 +29*r, where r=0, 1, 2, . . . , 31, and k is the minimum of values 31 and ?b/29!+1.