Fast Local Scale Control Based Blur Edge Detection

Abstract

Extracting objects with blurred edges from inhomogeneous background in the natural images is still under exploring. To locate blurred edges, a fault-tolerant method is adopted in the algorithm, to establish the unique, locally calculable minimum reliable scale for each pixel of the image. By simplifying the process of calculating the second derivation for each pixel, the convolution mask along its gradient direction is used. The calculation complexity of second derivation calculation in the gradient direction is reduced by combining the local scale control with LoG algorithm. Different algorithms are roughly benchmarked in the experiments among three different kinds of typical blurred images. The results show that the proposed algorithm localizes and extracts the blurred edges precisely, moreover it runs fast enough to perform some kind of real-time applications.