Combined spatial and transform domain analysis for rectangle detection

We describe and compare methods for detecting rectangles in images using Hough and Radon transforms. Locating rectangles in a new image involves an alternating scheme of transform peak extraction and peak filtering. During the step of peak extraction, we apply the hough/radon transform on the target image and extract peaks (corresponding to line segments) from the transformed image. We filter these peaks based on certain geometric constraints in the transform and spatial constraints on the coordinate (or image) domain such that every set of 4 filtered peaks correspond to a rectangle in the image. We explore the effect of model parameters on system performance and show that the proposed methods achieves good accuracy for rectangle detection on several synthetic and real datasets.