Geistiges Eigentum Herausforderung Durchsetzung
Die „Durchsetzung" hat sich im Recht des geistigen Eigentums zu einem ebenso zentralen wie vielschichtigen Begriff entwickelt. Der Sammelband beleuchtet die unterschiedlichen Aspekte des Begriffs, analysiert Entwicklungstendenzen, diskutiert gelöste und u
- PDF / 963,568 Bytes
- 9 Pages / 430 x 660 pts Page_size
- 77 Downloads / 193 Views
Abstract. Signals and images often suffer from blurring or point spreading with unknown filter or point spread function. Most existing blind deconvolution and deblurring methods require good knowledge about both the signal and the filter and the performance depends on the amount of prior information regarding the blurring function and signal. Often an iterative procedure is required for estimating the blurring function such as the Richardson-Lucy method and is computational complex and expensive and sometime unstable. In this paper a blind signal deconvolution and deblurring method is proposed based on an ICA measure as well as a simple genetic algorithm. The method is simple and does not require any priori knowledge regarding the signal and the blurring function. Experimental results are presented and compared with some existing methods.
1 Introduction A long-standing problem in image restoration is to reconstruct from a blurred and/or noisy image with as little as possible a priori knowledge of the original image, blurring function and the nature of added noise. The blurring degradation may be due to misfocus, motion or atmospheric turbulence. The resultant image should be as close to the original image as possible. Some related past efforts can be found in [1,6,7]. Here the motivation is to come up an effective and efficient method to perform blind deblurring. The degraded system can be represented by a general block diagram shown in Fig.1. Most blurring models rely on a standard model of a shift invariant kernel and additive noise, which mathematically can be represented as,
x (n1 , n2 ) = f (n1 , n2 ) ⊗ b( n1 , n2 ) + η ( n1 , n 2 )
(1)
where ⊗ is the convolution operator. The degraded image x(n1,n2) is the result of convoluting the original image f(n1,n2) with a point spread function (PSF) b(n1,n2) and then adding noise η(n1,n2). Various techniques have been applied to recover the actual image. Traditional approaches like median-based approaches are inadequate and limited. The ideal approach to deblurring is to directly inverse the PSF that degrades the image (same is true with the presence of noise). For the case when there is no or little noise, the direct E. Corchado et al. (Eds.): IDEAL 2006, LNCS 4224, pp. 595 – 603, 2006. © Springer-Verlag Berlin Heidelberg 2006
596
H. Yin and I. Hussain
η
f
Blurring Filter
+
x
Fig. 1. Blurring model. x is the corrupted or observed image, f is the original image and η is added noise.
inverse filter can be done easily in spectral (frequency) domain, where the convolution process becomes multiplication, F (ω1 , ω2 ) =
X (ω1 , ω2 ) B(ω1 , ω2 )
(2)
In most cases the PSF is not available. However, there are certain situations in which one may make certain valid assumption on the PSF or at least some properties. For example, if the blurring in the image is due to linear movement of the scene or camera during exposure, the PSF is like a sinc function, as shown in Fig. 2 (a). So the most straightforward approach is to recover the image through deconvolution wit
Data Loading...
