A blob detector in color images

Abstract

In this paper, we propose a novel and efficient framework for the extension of multi-scale blob detection for the color domain to prevent information loss due to gray scale transformation and to allow us to exploit the photometric information. The framework also works well in gray scale space, it is mostly composed of i) a weighted multi-scale blob detector using a hybrid Laplacian and determinant of the Hessian operator, ii) a blob filter, includes a color-based Forstner operator for roundness calculation and a hue-based color histogram. The experimental results on stamp images in blob detection demonstrate the effectiveness of our proposed detector.