Existing methods for extracting dominant textures are always complicated and very expensive, e. g. more than 10 minutes are often required to handle an image sample. This paper proposes an easy and fast method, being able to extract dominant textures from an image sample in less than 1 second, and its results are comparable with those of existing methods. The new method is based on the key observation that dominant textures often take up a larger portion in the image sample, and their color features appear frequently. Thus, with multiscale histograms, we may find the color features with high frequencies, and have their corresponding pixels belonged to the dominant textures. We adopt the HIS color model and re-sample the image sample in a few resolutions, and perform the investigation on them respectively, by which the pixels often regarded as belonged to dominant textures will be finally taken to form the dominant textures.
