A new methodology is presented for the computation of global illumination using structured sampling. Analytical/numerical solutions for illumination are developed for simple lighting configurations. These solutions are subsequently used to generate accurate reference images. The structured sampling solution for global illumination is then discussed, comprising sample placement for illumination calculation, reconstruction for light transfer and finally resampling and filtering of illumination samples for display. A first approximation to this technique is presented using a priori placement of samples, irregular polygon reflectors, grid resampling and a conical filter for display. The new algorithm is evaluated for image quality, and compared to the traditional radiosity-based approach. These first results show that the structured sampling solution yields significant computational savings while maintaining high image quality.