The dyadic shift-based image processing uses the address permutation of the individual pixels or the permutation of their gray levels or both by means of the XOR operation. It is properly described within the framework of the two-dimensional Walsh functions and Walsh transforms. The dyadic shift procedures allow the transformation of the original image and its posterior recovery without information loss by simply performing further permutations. In addition, comparison of particular features of different images can be achieved by applying dyadic correlations, which in turn can also take advantage of the Walsh transforms. Therefore, the dyadic shift-based image processing has potential applications on image encryption, image enhancement and filtering and pattern recognition among others.