We present a method to significantly reduce large-scale photometric variations seriously affecting imaging data from the Wide Field Imager (WFI). The primary source for these gradients is non-uniform illumination, which cannot be corrected by standard flatfielding techniques. Comparison of our observations with well-calibrated multi-colour photometry from the SDSS enabled us to characterize and quantify these variations and finally to model them using a second-order polynomial. Application of the model to our observations and an independent dataset consistently reduced the large-scale gradients and thus provides a generally valid and simple tool for improving WFI photometry.