This paper proposes an analytic design method for a class of 2D recursive filters, namely with directional and square-shaped frequency response. The design technique is based on a specified 1D digital prototype filter and a complex frequency transformation which is determined using various rational approximations. Several design examples are presented for given specifications. The resulted filters are efficient and adjustable, being at the same time of low complexity and relatively high selectivity. These filters may have useful applications in image processing, like detecting lines with a given orientation from an image, as shown through simulation results on test images. The novelty of the proposed method consists in deriving the transfer function of the desired 2D filter only by applying successive frequency transformations to a prototype filter. Its advantages over most state-of-the-art design methods, which generally use global optimization numerical algorithms, is the filter tunability, and also its versatility; since the filter parameters, namely orientation and selectivity appear explicitly in the 2D filter matrices, they result directly for any specified parameters, therefore the design procedure does not need to be resumed every time from the start.