<p>In computer vision, the pictorial structure model represents an object in an image by parts that are arranged in a deformable configuration. Each part describes an object's local photometric appearance, and the configuration encodes the global geometric layout. This model has been very successful in recent object recognition systems.</p><p>We extend the pictorial structure model in three aspects. First, when the model contains only a single part, we develop new methods ranging from regularized subwindow search, nested window search, to twisted window search, for handling richer priors and more flexible shapes. Second, we develop the notion of a weak pictorial structure, as opposed to the strong one, for the characterization of a loose geometric layout in a rotationally invariant way. Third, we develop nested models to encode topological inclusion relations between parts to represent richer patterns.</p><p>We show that all the extended models can be efficiently matched to images by using dynamic programming and variants of the generalized distance transform, which computes the lower envelope of transformed cones on a dense image grid. This transform turns out to be important for a wide variety of computer vision tasks and often accelerates the computation at hand by an order of magnitude. We demonstrate improved results in either quality or speed, and sometimes both, in object matching, saliency measure, online and offline tracking, object localization and recognition.</p> / Dissertation

This thesis suggests a modification to the popular Scale Invariant Feature Transform (SIFT) algorithm (Lowe, 2004) often used in photogrammetry and computer vision to find features in images for measurement. The SIFT algorithm works by first detecting points in images at different scales and sizes. It then refines the position of the found points. The algorithm creates a descriptor of the point based on the region around the point. Finally the points can be matched against other points in different images using the descriptor. The suggested modification is built upon a paper by Förstner and Gülch (1987) where a method for performing a subwindow localization is presented. In this thesis the keypoints detected by the SIFT algorithm are modified on the subwindow level in order to improve the robustness with respect to the selected window position. Several different methods of tweaking the suggested modification and the SIFT algorithm were tested. The methods were evaluated on two different test cases. The first used a camera calibration software to compare accuracy of keypoints by looking at the residuals of the calibration. The other test involved creating a point cloud of images of a planar surface, evaluating the results by looking at the standard deviation in keypointoffset from the plane.The results show that neither test gave evidence that the proposed modification was an improvement. It was found that the algorithm had problems with oblique projections of circles, i.e. ellipses. Therefore there is potentialto use homography in special cases to circumvent this problem and get better precision. Furthermore tests involving more lines and intersections in the test images should be performed before this suggested modificationcan be completely discarded.