The use of ICA (independent component analysis) for the construction of filters for lighting invariant face recognition will be investigated. ICA is used to provide filters which are applied as a pre-processing step to a low dimensional PCA subspace representation of the databases. Test faces imaged under varying illumination from a face database are classified using a support vector classifier. The ICA pre-filter recognition results are compared against those using LoG (Laplacian of Gaussian) filter of various spatial resolutions and no pre-filtering. The ICA pre-filters are shown to be very effective at selectively reducing the effect of illumination variance in object and face recognition without the need for tuning the filters to the orientations and spatial resolutions present in the images.