Matching pursuit‐based shape representation and recognition using scale‐space

In this paper, we propose an analytical low‐level representation of images, obtained by a decomposition process, namely the matching pursuit (MP) algorithm, as a new way of describing objects through a general continuous description using an affine invariant dictionary of basis function (BFs). This description is used to recognize multiple objects in images. In the learning phase, a template object is decomposed, and the extracted subset of BFs, called meta‐atom, gives the description of the object. This description is then naturally extended into the linear scale‐space using the definition of our BFs, and thus providing a more general representation of the object. We use this enhanced description as a predefined dictionary of the object to conduct an MP‐based shape recognition task into the linear scale‐space. The introduction of the scale‐space approach improves the robustness of our method: we avoid local minima issues encountered when minimizing a nonconvex energy function. We show results for the detection of complex synthetic shapes, as well as real world (aerial and medical)