Retrieving images in a database based on user specified colors is a popular low-level retrieval technique. However, the available systems today do not easily allow for a user or a specified query to tag certain colors as unwanted in the query result to ultimately be excluded in the query. Specifically, color histogram techniques do not allow for a direct approach to excluding colors and would require a separate query stage to filter out images containing unwanted colors. In this paper we present our vector-based scheme to image retrieval using a multidimensional query space which naturally accepts the exclusion of specified colors in the overall similarity measure.