A method is presented for kNN search of images with the aid of two-stage filtering which is an extension of an improved single-stage filtering method presented by Seidl et al. The two-stage filtering method is applied to image retrieval. Three kinds of feature vectors are examined for color-based representation of image contents, those are color histograms, block-partitioned histograms and pixel color vectors. It is verified that retrieval time is shortened by extension of filtering to two-stage for all of these feature vectors.