1 Ultrasound Image Denoising by Spatially Varying Frequency Compounding Yael Erez , Yoav Y. Schechner , and Dan Adam Dept. Elect. Eng. Technion – Israel Institute of Technology
7 Ultrasound Problems Transmitter Receiver Blurring Speckle noise Attenuatio n System noise Erez, Schechner & Adam , Proc. DAGM 2006 Radial axis Lateral axis
Previous Work 70 s Wiener filter 80 s Compounding (frequency & spatial) Not using noise statistics Space invariant 86 Weighted median filter (Mcdicken et al. ) 89 Local frequency diversity (Forsberg et al. ) Smoothing Anisotropic handling (Perona and Malik) Not diffusion attenuation 90 95 Late 90 s 01, 04 Non-linear Gaussian filters (Aurich) Harmonic imaging Low signal Wavelets (Insana et al, Loi et al. )
8 Image Formation probe Received signal Erez, Schechner & Adam , Proc. DAGM 2006 Velocity of acoustic wave in tissue
Image Formation 9 Sector Probe Radial axis Sweeping beam Lateral axis Erez, Schechner & Adam , Proc. DAGM 2006
10 Lateral PSF D D Low acoustic freq High acoustic freq (? ) r e . = q fre h Hig ett b Erez, Schechner & Adam , Proc. DAGM 2006
11 Attenuation probe a object r distance L freq. ow Erez, Schechner & Adam , Proc. DAGM 2006 = r (? ) bette
Speckle Noise Low acoustic freq 15 High acoustic freq Erez, Schechner & Adam , Proc. DAGM 2006
16 Wave phenomenon Object blur: Wave interference as if no interference Speckle Noise Erez, Schechner & Adam , Proc. DAGM 2006
PSF 17 D D Low acoustic freq High acoustic freq Depends on: • Radial distance • Acoustic frequency Erez, Schechner & Adam , Proc. DAGM 2006
Measuring Noise Statistics 1 0. 8 White noise 18 r = 7 cm r = 11 cm r = 15 cm 0. 6 0. 4 0. 2 0 -2 -1 0 1 Radial lag (mm) 2 Erez, Schechner & Adam , Proc. DAGM 2006
19 Standard Pre-Processing RF line Sampling Time gain compensation Envelope detection Dynamic range compression
Speckle Noise 20 = log operation Iinear noise Erez, Schechner & Adam , Proc. DAGM 2006
21 … … … Model correlated noise !!! Erez, Schechner & Adam , Proc. DAGM 2006
22 H x + n … = … … y Stochastic Reconstruction Erez, Schechner & Adam , Proc. DAGM 2006
Best Linear Unbiased Estimator Considering noise statistics Erez, Schechner & Adam , Proc. DAGM 2006 23
![Input: Dual Acoustic Frequency Radial distance [cm] 5 Low acoustic freq 6 7 8](https://present5.com/presentation/7ad42da01b65c06477801f74735b7b9e/image-17.jpg "Input: Dual Acoustic Frequency Radial distance [cm] 5 Low acoustic freq 6 7 8")
Input: Dual Acoustic Frequency Radial distance [cm] 5 Low acoustic freq 6 7 8 9 10 11 Erez, Schechner & Adam , Proc. DAGM 2006 High acoustic freq 24
![Stochastic Freq. Compounding Radial distance [cm] 5 Arithmetic mean 6 7 8 9 10](https://present5.com/presentation/7ad42da01b65c06477801f74735b7b9e/image-18.jpg "Stochastic Freq. Compounding Radial distance [cm] 5 Arithmetic mean 6 7 8 9 10")
Stochastic Freq. Compounding Radial distance [cm] 5 Arithmetic mean 6 7 8 9 10 11 Erez, Schechner & Adam , Proc. DAGM 2006 Stochastic reconstruction 25
Скачать презентацию 1 Ultrasound Image Denoising by Spatially Varying Frequency
7ad42da01b65c06477801f74735b7b9e.ppt