The Study Of Defocusing Iris Image Restoration Based On Curvelet Transform

With the coming of information era and the continuous development ofscience and technology, the importance of information security has graduallyhighlighted. In the process of constant research and demonstration, biometricidentification technology began to more and more be taken seriously. Irisrecognition, as a kind of biological recognition technology, has someadvantages such as strong stability and high recognition rate. Iris recognitiontechnology has become the research hotspot of biometric field.In1993, J.Daugman proposed a high performance of iris recognitionalgorithm. But the application of iris recognition technology is still in itsinfancy. But the precondition of recognition is getting the clear iris image. Butin fact, the iris image that we gained is often defocusing. The blurry iris imagewill bring very great difficulty in iris recognition. Therefore, how to get cleariris image becomes an important subject in the study of iris recognition. Atpresent, the solution to the blurried iris image basically has two kinds: one iswith automatic zoom image acquisition equipment. But the equipment isexpensive. Of course, it has increased the cost. Another kind is with thefixed-focus image acquisition equipment. So the study of the defocused irisimage restoration has the very good practical significance. In this paper, theproposed method belongs to the second kind.This paper analysis the possible causes of defocusing iris images, such aslight, relative movement, the distance, equipment size.Whether one or more ofthe above-mentioned factors, the result is reducing the optical depth of fieldand making the image defocused. In most cases, the defocusing distance isalways unpredictable, which brings the difficulties to the image restoration.In this paper, we used the power spectral analysis method to established therelationship between the out-of-focus radius and the spectrum of image. Basedon this relation,we proposed the focusing evaluation function of the image. It isto say that we can estimate the out-of-focus radius according to the imagepower spectrum t. Then we could estimate the point spread function (PSF).In this paper, we study a image restoration method based on hybrid transformdomain. Because the texture feature in the iris image is important for thefollow-up recognition algorithm, we choose curvelet transform to process thetexture image. The proposed algorithm takes Fourier domain and curveletdomain contraction at the same time. The algorithm exploits the Fouriertransform’s sparse representation of the colored noise and curvelet transform’ssparse representation of image texture and edge character.Finally, the iris image restoration algorithm is as follows: firstly, by focusingevaluation function judges whether the input iris image is out of focus. Thenthe point spread function is estimated. And the iris image is processed withdiscrete Fourier transform. Then contract in Fourier domain to get thepreliminary result. At the same time reduce the noise variance. The preliminaryresult is processed with discrete curvelet transform, denoising while retainingmost of the image texture information. Then using curvelet inversetransformation processes the iris image. Experiments show that this algorithmget a good result.