LCE/7.5.

1/RC 01

TEACHING NOTES
Department: ELECTRONICS & COMMUNICATION ENGINEERING Unit: VI Date: Topic name: Introduction to Degradation Model No. of marks allotted by JNTUK: Books referred: 01. Digital Image Processing by R C Gonzalez and R E Woods 02. www.wikipedia.org 03. www.google.com Introduction: The purpose of image restoration is to compensate for or undo defects which degrade an image. Degradation comes in many forms such as motion blur, noise, and camera misfocus. In cases like motion blur, it is possible to come up with a very good estimate of the actual blurring function and undo the blur to restore the original image. In cases where the image is corrupted by noise, the best we may hope to do is to compensate for the degradation it caused. In this project, we will introduce and implement several of the methods used in the image processing world to restore images. Degradation model: The block diagram for our general degradation model is,

Where g is the corrupted image obtained by passing the original image f through a low pass filter (blurring function) b and adding noise to it. We present four different ways of restoring the image. Inverse Filter: In this method we look at an image assuming a known blurring function. We will see that restoration is good when noise is not present and not so good when it is. Weiner Filtering: In this section we implement image restoration using wiener filtering, which provides us with the optimal trade-off between de-noising and inverse filtering. We will see that the result is in general better than with straight inverse filtering. Wavelet Restoration: We implement three wavelet based algorithms to restore the image. Blind De-convolution: In this method, we assume nothing about the image. We do not have any information about the blurring function or on the additive noise. We will see that restoring an image when we know nothing about it is very hard. Faculty/Date: Page 1 of 4 HOD/Date:

LCE/7.5.1/RC 01

TEACHING NOTES
Department: ELECTRONICS & COMMUNICATION ENGINEERING Unit: VI Date: Topic name: Inverse Filtering No. of marks allotted by JNTUK: Books referred: 01. Digital Image Processing by R C Gonzalez and R E Woods 02. www.wikipedia.org 03. www.google.com Inverse Filtering: If we know of or can create a good model of the blurring function that corrupted an image the quickest and easiest way to restore that is by inverse filtering. Unfortunately, since the inverse filter is a form of high pass filter, inverse filtering responds very badly to any noise that is present in the image because noise tends to be high frequency. In this s ection, we explore two methods of inverse filtering a thresholding method and an iterative method. Method-1 (Thresholding): We can model a blurred image by,

Where f is the original image, b is some kind of a low pass filter and g is our blurred image. So to get back the original image, we would just have to convolve our blurred function with some kind of a high pass filter,

In the ideal case, we would just invert all the elements of B to get a high pass filter. However, notice that a lot of the elements in B have values either at zero or very close to it. Inverting these elements would give us either infinities or some extremely high values. In order to avoid these values, we will need to set some sort of a threshold on the inverted element. So instead of making a full inverse out of B, we can an almost full inverse by

So the higher we set , the closer H is to the full inverse filter. Method-2 (Iterative procedure): The idea behind the iterative procedure is to make some initial guess of f based on g and to update that guess after every iteration. The procedure is,

Where f is an initial guess based on g. if our fk is a good guess, eventually fk convolved with b will be close to g. when that happens the second term in the fk+1 equation will disappear and fk and fk+1 will converge. is our convergence factor and it lets us determine how fast fk and fk+1 converge. If we take both of the above equations to the frequency domain, we get

Faculty/Date: Page 2 of 4

HOD/Date:

Dep t ent: ELECTRONICS & COMMUNICATION ENGINEERING Unit: VI Date: Topi name: Leas Mean S are F ers No of marks allotted b JNTUK: Books referred: 01 Digital Image Processing by R C Gonzalez and R E Woods 02 ikipedia.org 03 www.google.com

Solving for Fk rec rsively we get

So if (1- B( 1 2 k+1 goes to zero as k goes to infinity we would get the result as obtained by the inverse filter. In general, this method will not give the e act same results as inverse filtering, but can be less sensitive to noise in some cases. Least Mean Square Filters: Least mean s uares (LMS) algorithms are used in adaptive filters to find the filter coefficients that relate to producing the least mean s uares of the error signal ( d n b w n he desired and the a tua signa ). It is a stochastic gradient descent method in that the filter is only adapted based on the error at the current time. It was invented in 1960 by Stanford University professor Bernard Widrow and his first Ph.D. student, Ted Hoff.

Most linear adaptive filtering problems can be formulated using the block diagram above. That is, an unknown system is to be identified and the adaptive filter attempts to adapt the filter to make it as close as possible to , while using only observable signals x(n), d(n) and e(n); but y(n), v(n) and h(n) are not directly observable. Its solution is closely related to the Wiener filter. The idea behind LMS filters is to use the method of steepest descent to find a coefficient vector h(n) which minimizes a cost function. We start the discussion by defining the cost function as

Faculty Date:

Page 3

8 55 8 5 57 565 443

HOD/Date: LCE/7.5.1/RC 01

0 '

"

TEA
!     @ 

OTE

751

01

 

'



$ %" %%% $ $ )) & 1 ' 9

Department: ELECTRONICS & COMMUNICATION ENGINEERING Unit: VI Date: Topic name: Constrained Least S uares Restoration No. of marks allotted b JNTUK: Books referred: 01. Digital Image Processing by R C Gonzalez and R E Woods 02. www.wikipedia.org 03. www.google.com Where e(n) is defined in the block diagram section of the general adaptive fiter and E{.} l denotes the expected value. Applying the steepest descent method means to take the partial derivatives with respect to the individual entries of the filter coefficient vector Where is the gradient operator with it follows,

Now,

is a vector which points towards the steepest ascent of the cost function. To .

find the minimum of the cost function we need to take a step in the opposite direction of To express that in mathematical terms

Where is the step size. That means we have found a se u ential update algorithm which minimizes the cost function. Unfortunately, this algorithm is not realizable until we know . Constrained Least Squares Restoration: Image restoration using the constrained least-s uares (CLS) method theoretically adapts to the image being processed. In addition, it only re uires knowing the modulation transfer function of the imaging system when applied to nuclear medicine images. Prompted by these observations, a systematic evaluation of the effects of the form of the "coarseness function" [C (f)] used by the CLS method has been conducted. Nine C(f)'s are evaluated using an observer preference and a normalized mean-s uared error (NMSE) criterion. This evaluation is conducted for three modulation transfer functions and a wide range of count levels. The results of the subjective studies support using the form of C(f) which has been most widely employed in previous studies, i.e., the form designed to minimize the energy in the second derivative of the restored image. A different form of C(f) is generally found to be optimal by the mean-s uared error criterion. The CLS method is then compared to: (1) no processing, (2) count-dependent smoothing, and (3) count-dependent Metz restoration. When evaluated using objective measurements of error and contrast, the CLS method is found to be slightly inferior to the best method, Metz restoration. However, CLS restoration is found to be e ual to or better than the other methods when judged by the results of observer preference studies. Faculty/Date:

HOD/Date:

Page 4 of 4

TEA

OTE

E
and

C DC B