电子转发女学光电科学与工程学院 SCHOOL OF OPTOELECTRONIC SCIENCE AND ENGINEERING OF UESTC 1.Noise Probability Density Functions p(z) ■Uniform noise 1 b-a Uniform 1/(b-a),ifa≤z≤b p(z)= 0, otherwise where =(a+b)/2,02=(b-a)'/12 e b p(z) Impulse (Salt Pepper)noise Pb Impulse Pa:ifz=a Pa p(2)=了P6,ifz=b 0,otherwise a b 2
◼ Uniform noise z a b = + ( ) / 2, ( ) 2 2 = −b a /12 1/( ),if ( ) 0, otherwise b a a z b p z − = ≤ ≤ 1. Noise Probability Density Functions where ◼ Impulse (Salt & Pepper) noise , ( ) , 0, otherwise a b P if z a p z P if z b = = =
电子转发女学光电科学与工程学院 SCHOOL OF OPTOELECTRONIC SCIENCE AND ENGINEERING OF UESTC 1.Noise Probability Density Functions P() p() p(z) 1 0.607 2 V2TG K Gaussian Rayleigh Gamma 0.607 V2TO a(b-1)- e(h-) b-1！■ +o Q b (b-1)/a a+ V2 P(z)川 p()川 p(Z) 1 Pa a Exponential b-a Uniform Impulse a b Z a abc de f FIGURE 5.2 Some important probability density functions
1. Noise Probability Density Functions
Example-noisy images and their histograms p(z) p(z) p() 1 √2ru o.v7 Gaussian Rayleigh Gamma 0.607 V2To K= a(b-1-1 e-(b-1) (b-1)1 元-u元+u b (b-1)/a a+
Example – noisy images and their histograms
Example-noisy images and their histograms ■matlab codes J=imnoise(I,type, parameters) 1) gaussian 2 localvar 3 poisson salt pepper pP(3)川 p() p() 1 Po 0 5) speckle Exponential b-a Uniform Impulse Pa DEMO a 6 6
Example – noisy images and their histograms DEMO 1 3 4 5 2 gaussian localvar poisson salt & pepper speckle J = imnoise(I,type, parameters) ◼ matlab codes
电子科线女学光电科学与工程学院 SCHOOL OF OPTOELECTRONIC SCIENCE AND ENGINEERING OF UESTC 2.Periodic Noise a FIGURE 5.5 (a)Image corrupted by sin(2π4ox+2πvoy) sinusoidal noise. (b)Spectrum (each pair of conjugate impulses corresponds to one sine wave). (Original image courtesy of NASA.) j56(u+M4,+M,） -6(u-M4,v-Nvo)]
2. Periodic Noise 0 0 sin (2 x+2 y) v ( ) ( ) 0 0 0 0 1 [ , 2 , ] j M v Nv M v Nv + + − − −