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

摘要

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