Keypoint detection is a basic step in many computer vision algorithms aimed at recognition of objects, automatic navigation and analysis of biomedical images. Successful implementation of higher level image analysis tasks, however, is conditioned by reliable detection of characteristic image local regions termed keypoints. A large number of keypoint detection algorithms has been proposed and verified. In this paper we discuss the most important keypoint detection algorithms. The main part of this work is devoted to description of a keypoint detection algorithm we propose that incorporates depth information computed from stereovision cameras or other depth sensing devices. It is shown that filtering out keypoints that are context dependent, e.g. located at boundaries of objects can improve the matching performance of the keypoints which is the basis for object recognition tasks. This improvement is shown quantitatively by comparing the proposed algorithm to the widely accepted SIFT keypoint detector algorithm. Our study is motivated by a development of a system aimed at aiding the visually impaired in space perception and object identification.
The task of generating fast and accurate three-dimensional (3D) models of objects or scenes through a sequence of non-calibrated images is an active field of research. The recent development in algorithm optimization has resulted in many automatic solutions that can provide an accurate 3D model from texture-full objects. Structure-from-motion (SfM) is an image-based method that uses discriminative point-based feature identifier, such as SIFT, to locate feature points in the images. This method faces difficulties when presented with the objects made of homogenous or texture-less surfaces. To reconstruct such surfaces a well-known technique is to apply an artificial variety by covering the surface with a random texture pattern prior to the image capturing process. In this work, we designed three series of image patterns which are tested based on the contrast and density ratio which increases from the first to the last pattern within the same series. The performance of the patterns is evaluated by reconstructing the surface of a texture-less object and comparing it with the true data. Using the best-found patterns from the experiments, a 3D model of a Moai statue is reconstructed. The experimental results demonstrate that the density and structure of a pattern highly affects the quality of the reconstruction.