KOTROPOULOS C. and PITAS I., Optimum nonlinear signal detection and estimation in the presence of ultrasonic speckle, Ultrasonic Imaging 14, 249-275, (1992).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....tation algoritms e#ciently. Cons#Hhj tly,s peckle reduction filteringis an importantas ect forvisI4H:hj:FC and sndh4 tation of ultrashj4 images Speckleis modelled in a multiplicative Rayleig nois t atdegrades ultrases images by concealing finesh4I:CXhj and reducing t esXIHL tonois ratio (SNR) [9]. Kotropoulos andPitas [9] derived a maximum likeli oodes:#HHh4 for ultrash:4 stras corrupted bynois: T e maximum of t e Rayleig dish:XI:Hhj provides a maximum likeli ood esS#SShj [9] T e maximum of t e probabilitydensI y function (PDF)is given by derivative of t e PDF wit res ect to zas ....

.... Cons#Hhj tly,s peckle reduction filteringis an importantas ect forvisI4H:hj:FC and sndh4 tation of ultrashj4 images Speckleis modelled in a multiplicative Rayleig nois t atdegrades ultrases images by concealing finesh4I:CXhj and reducing t esXIHL tonois ratio (SNR) 9] Kotropoulos andPitas [9] derived a maximum likeli oodes:#HHh4 for ultrash:4 stras corrupted bynois: T e maximum of t e Rayleig dish:XI:Hhj provides a maximum likeli ood esS#SShj [9] T e maximum of t e probabilitydensI y function (PDF)is given by derivative of t e PDF wit res ect to zas follows #p(z) #z =0=exp # z ....

[Article contains additional citation context not shown here]

C. Kotropoulos and I. Pitas, "Optimum nonlinear signal detection and estimation in the presence of ultrasonic speckle," Ultrasonic Imaging, vol.14, pp.249--275, 1992.

....and by the different image processing filters which might be used for better visualization of the echo signal. We assume that the image is corrupted by multiplicative noise (Gaussian with a zero mean) and a logarithmic compression dominates resulting in a signal dependent noise model (see [76] and [73] for detailed explanation) ii) The kernel of the left ventricle (central portion of the left ventricle that resembles its shape) and the boundary area (where the contour should be located) belong to different gray scale intervals that might be overlapped, but the expected (nominal) ....

....noise, neither the median nor the mean can be considered as an optimal estimator for US images. Kontropoulos et al. 73] showed that the ML (maximum likelihood) estimator for displayed US image data (signal dependent noise model) closely resembles the L 2 mean which has been proven earlier [76] to be the ML estimator of the original signal in US B mode data (multiplicative Rayleigh distributed noise model) They designed signal adaptive ML filters for both models. Evans and Nixon [75] proposed that, alternatively, the maximum of the Rayleigh distribution provides a value (the ....

[Article contains additional citation context not shown here]

C. Kontropoulos and I. Pitas, "Optimum Nonlinear Signal Detection and Estimation in the Presence of Ultrasonic Speckle", Ultrasonic Imaging, Vol. 14, No. 3, pp. 249-275, July 1992.

....image data have undergone excessive manipulation (e.g. logarithmic compression, low and high pass filtering, postprocessing, etc. In the case of pure multiplicative Rayleigh speckle, it has been proved that the maximum likelihood (ML) estimator of the original (noiseless) signal is the L2 mean [16]. Furthermore, for signal dependent Gaussian speckle, it has been shown that the ML estimator closely resembles the L2 mean [6] These observations motivated us to modify the standard LVQ algorithm so that the reference vectors correspond to the L2 mean instead of the sample arithmetic mean. Such ....

KOTROPOULOS C. and PITAS I., Optimum nonlinear signal detection and estimation in the presence of ultrasonic speckle, Ultrasonic Imaging 14, 249-275, (1992).

No context found.

C. Kotropoulos, and I. Pitas, "Optimum nonlinear signal detection and estimation in the presence of ultrasonic speckle," Ultrasonic Imaging, vol. 14, no. 3, pp. 249--275, July 1992.

....remarks explain well enough why LVQ, despite its simplicity, is sufficient for a US image segmentation task. However, it should be noted that the arithmetic mean approximated by the basic LVQ, described so far, is not the best possible estimator of the mean level in a US image. It has been proved [12] that the maximum likelihood estimator of the original noiseless image is the L 2 mean [22] scaled by p 2 ) of the noisy observations x i ; i = 1; M , i.e. p 2 v u u t 1 M M X i=1 x 2 i : 3) This result leads us to consider a modification of the standard LVQ ....

....contrast, particularly in the background area, has been reduced, yielding a better discrimination of the lesion against the background. The corresponding filters coefficients are given in Fig. 5. Observe that, although the background filter resembles the Rayleigh optimum L filters computed in [12], the optimum L filter for the lesion area is very close to the median filter. This result is not unexpected in view of the Laplacian form of the histogram of the bright region shown in Fig. 2b. Some classical filtering operators The arithmetic mean and the median filters have also been used in ....

C. Kotropoulos and I. Pitas, "Optimum Nonlinear Signal Detection and Estimation in the Presence of Ultrasonic Speckle," Ultrasonic Imaging, vol. 14, 1992, pp. 249--275.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC