In this paper, a method to spot and recognize logos based on key-point matching is proposed. It is applied and tested on a document retrieval system. First, the pairs of matched key-points are estimated by the nearest neighbor matching rule based on the two nearest neighbors in SIFT descriptor space with Euclidean distance. Second, a post-filter with BRIEF descriptor space and hamming distance is used to re-filter the key-points which are rejected by the first step. Tested on a well-known benchmark database of real world documents containing logos Tobacco-800, our method performs an increase in the number of matched key-points of the method combined with BRIEF post-filter at the same accuracy level, and achieves a better performance than the state-of-the-art methods in the field of document retrieval.