Abstract: A video image, characterized by a plurality of pixels, is compressed by deriving at least one parameter based upon principal component analysis. An image value for each video image is then determined for the one parameter. The image value associated with each video image is compared to an image value associated with another video image to determine a difference. Any one of the image values of each video image and another video image is discarded in the event the difference is below a threshold. The remaining video images are then stored. Using this method of compression, it becomes possible to search a video database to find a match between an inquiring video scene with a video database comprising a plurality of video scenes with each video scene having a plurality of video images.