Spectral Residual Method Of Saliency Detection Based On The Two-Dimensional Fractional Fourier Transform Domain

Human’s function of notice selection enable a person to extract interested area from a image immediately. Saliency detection imitate the character of human vision to extract the saliency map and has widely application in many fields. Saliency detection methods could be divided into two different approach: bottom-up approach and top-down approach. Bottom-up approach is the core of saliency detection and the focus of this paper. As one of classic methods of frequency domain based saliency detection, spectral residual(SR) method has shown several advantages. However, it usually produces higher saliency values at object edges instead of generating maps that uniformly cover the whole object, which results from failing to exploit all the spatial frequency content of the original image.Several classical methods has been introduced. Aiming at the shortages of time domain based methods and frequency domain based methods, this paper introduce several recent improved methods and propose a method based on two-dimensional fractional Fourier transform(2D-FRFT), which improved spectral residual(SR) method.The two-dimensional fractional Fourier transform(2D-FRFT) is a generalized form of the traditional Fourier transform(FT) which can abstract more meaningful information of the image under certain conditions. Based on this property, we propose a new method which detects the salient region based on 2D-FRFT domain. This method takes full advantage of the overall and local information, which can solve the problems of the spectral residual method.Moreover, we also propose a new method based on Hough transform to refine the saliency map. We conduct experiments on three common used datasets: Bruce, MIT, MSRA. The proposed method is compared with several other saliency detection methods and shown to achieve superior result.