MINI PROJECT ON A NEW REFERENCE-FREE IMAGE QUALITY INDEX FOR BLUR ESTIMATION IN THE FREQUENCY DOMAIN DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING 1 BY M.V.S.RAMA RAJU(09591A0461) P.SAIDEEP(09591A0476) UNDER THE GUIDANCE OF K.PRABHU
MINI PROJECT ON A NEW REFERENCE-FREE IMAGE QUALITY INDEX FOR BLUR ESTIMATION IN THE FREQUENCY DOMAIN
DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING
1
BY
M.V.S.RAMA RAJU(09591A0461) P.SAIDEEP(09591A0476) UNDER THE GUIDANCE OF K.PRABHU
overview Introduction What is blur? Why blur occurs? Difference between noise and blur? Types of images Block diagram for blur estimation Math content Sample code Different types of filtering techniques Why estimation? Applications References Future scope Conclusion
2
IntroductionThis project presents a new reference-free
image quality index based on spectral analysis
The proposed method consists of adding blur to the test image and measuring its impact.
3
What is blur?Change in the position of pixels in an image
Original image Blurred image
4
How blur occurs ?Blur occurs due to handshake while taking a
picture in a cell cameraBlur occurs due to motion of objectsBlur occurs due to atmospheric conditions
5
Difference between noise and blur
Image with noise Image with blur
6
Types of imagesGray scale
Indexed
Binary
Rgb
Gray scale
R g b
0 or 1
R g b
7
Block diagram for blur estimation
+Blurred image
Fourier transform
Radial analysis
Blur index
Blurpsf(x,y)
o(x,y) +Blurred image +
noise
noisen(x,y)
8
r(x,y)
Math contThe discrete Fourier transform of 2D
F(u,v) is then transformed into polar coordinates represented by F(ω, θk).
An expression for the total radial energy is then obtained using the following:
9
Sample code %%% calculate the FFT for test and filterd images %%% Ft=fft2(I(:,:,k)); Ff=fft2(bimg(:,:,k)); for i=1:length(Ft) for j=1:length(Ft) ER(i,j)=abs(cart2pol(imag(Ft(i,j)),real(Ft(i,j)))); ERf(i,j)=abs(cart2pol(imag(Ff(i,j)),real(Ff(i,j)))); end end %%%% calculate the BI %%%% wmax=max(max(ER)); BI(:,:,k)=log((1/wmax)*sum(sum(abs(ER-ERf)))); end figure,imshow(dimg);title('Degraded image'); figure,imshow(bimg);title('Blurred image'); bb=(BI(:,:,1)+BI(:,:,2)+BI(:,:,3))/3; disp(['The Blur Index of the image is ',num2str(bb)])
10
Different types of filtering techniquesInverse filteringWiener filteringRegularised filtering
11
WHY ESTIMATION?
Blurred image Estimation of original image 12
Applications
Cameras at traffic signalsSatellite imagesComputer animation photography
13
Future scope:
•Also in future scope, this algorithm can be very useful for the real time video image processing by improving the live feature extraction parameters. •The proposed technique can enhance the future of the multimedia digital image processing by means of super resolution.
conclusion• An efficient no-reference blur quality index is
proposed. The method is not based on edge detection as most existing methods.
• Instead, we use a basic radial analysis in the frequency domain to measure the impact of blur added to the original image.
• The obtained results in terms of correlation with the subjective tests prove the efficiency of the proposed method.
15
References:
Freeman, W.T.; Pasztor, E.C.. Learning low-level vision, CVPR 1999, 182-1189 vol.2
William T. Freeman, Thouis R. Jones, and Egon C. Pasztor, Example-based super-resolution, IEEE Computer Graphics and Applications, March/April, 2002
Baker, S.; Kanade, T. Hallucinating faces. Automatic Face Gesture Recognition, 2000, 83-88.
Ce Liu; Heung-Yeung Shum; Chang-Shui Zhang. A two-step approach to hallucinating faces: global parametric model and local nonparametric model. CVPR 2001. I-192-8.
16
Queries
17