In this paper we propose a novel method to reduce the computational complexity of Circular Hough Transform (CHT). This Transform is used in identifying radial shapes in binary images. Hence, a novel adaptive algorithm is presented that reduces the average accumulation storage that is needed during voting process of CHT. The key idea is to use a variable size voting array. For this purpose, our method calculates the Signature Curve of the gradient of the edge points of radial shapes to find the center and radius of the probable circles, thus removing the effect of inconstant background luminance, specifically in presence of the noise. Experimental assessments show that the presented algorithm can considerably improve the identification precision, and decrease the computational burden and storage required.
