Scale detection via keypoint density maps in regular or near-regular textures

In this paper we propose a new method to detect the global scale of images with regular, near regular, orhomogenous textures. We define texture ‘‘scale’’ as the size of the basic elements (texels or textons) thatmost frequently occur into the image. We study the distribution of the interest points into the image, atdifferent scale, by using our Keypoint Density Maps (KDMs) tool. A ‘‘mode’’ vector is built computing themost frequent values (modes) of the KDMs, at different scales. We observed that the mode vector is quasilinear with the scale. The mode vector is properly subsampled, depending on the scale of observation, andcompared with a linear model. Texture scale is estimated as the one which minimizes an error functionbetween the related subsampled vector and the linear model. Results, compared with a state of the artmethod, are very encouraging.