(3 intermediate revisions by the same user not shown)
Line 49: Line 49:
 
</math>
 
</math>
  
Therefore the filter can be separated into two 1-D filters.
+
Therefore the filter can be decomposed to two 1-D filters.
  
<math>h_1[m] = -\frac{1}{4}\delta[m+1] + \delta[m] -\frac{1}{4}\delta[m-1]</math>  
+
<math>h_1[m] = \frac{1}{4}(\delta[m+1] + 2\delta[m] +\delta[m-1])</math>
 +
 
 +
<math>h_2[n] = \frac{1}{4}(\delta[n+1] + 2\delta[n] +\delta[n-1])</math>  
  
<math>h_2[n] = \frac{1}{2}\delta[n+1] + \delta[n] +\frac{1}{2}\delta[n-1]</math>
 
  
 
c) Compute the DSFT H(u,v) of this filter. Sketch the plot of H(u,0). Sketch the plot of H(0,v). What are the characteristics of this filter (low-pass, band-pass, or high-pass)?
 
c) Compute the DSFT H(u,v) of this filter. Sketch the plot of H(u,0). Sketch the plot of H(0,v). What are the characteristics of this filter (low-pass, band-pass, or high-pass)?
  
  
<math>H_1(\mu) = DTFT\{h_1[m]\} = -\frac{1}{4}e^{-j\mu(-1)} + e^{-j\mu(0)} -\frac{1}{4}e^{-j\mu(1)} = 1-\frac{1}{2}cos\mu</math>
+
<math>H_1(\mu) = DTFT\{h_1[m]\} = \frac{1}{4}(e^{j\mu} + 2 + e^{-j\mu}) = \frac{1}{2}(1 + cos\mu)</math>
  
<math>H_2(\nu) = DTFT\{h_2[n]\} = \frac{1}{2}e^{-j\nu(-1)} + e^{-j\nu(0)} +\frac{1}{2}e^{-j\nu(1)} = 1+cos\nu</math>
+
<math>H_2(\nu) = DTFT\{h_2[n]\} = \frac{1}{4}(e^{j\nu} + 2 + e^{-j\nu}) = \frac{1}{2}(1 + cos\nu)</math>
  
 
Using the separability,
 
Using the separability,
  
<math>H(\mu, \nu) = DSFT\{ h[m,n]\} = H_1(\mu)\cdot H_2(\nu) = (1-\frac{1}{2}cos\mu)(1+cos\nu)</math>
+
<math>H(\mu, \nu) = DSFT\{ h[m,n]\} = H_1(\mu)\cdot H_2(\nu) = \frac{1}{4}(1 + cos\mu)(1 + cos\nu)</math>
  
<math>H(\mu, 0) = 2(1-\frac{1}{2}cos\mu)</math>
+
<math>H(\mu, 0) = \frac{1}{2}(1 + cos\mu)</math>
  
[[Image:HW11_prob3_1.jpg]]
+
[[Image:HW11_prob1_1.jpg]]
  
So, <math>H(\mu, 0)</math> is a high-pass filter.
+
So, <math>H(\mu, 0)</math> is a low-pass filter.
  
  
 
<math>H(0, \nu) = \frac{1}{2}(1+cos\nu)</math>
 
<math>H(0, \nu) = \frac{1}{2}(1+cos\nu)</math>
  
[[Image:HW11_prob3_2.jpg]]
+
[[Image:HW11_prob1_2.jpg]]
  
 
So, <math>H(0, \nu)</math> is a low-pass filter.
 
So, <math>H(0, \nu)</math> is a low-pass filter.
Line 97: Line 98:
 
\end{array}
 
\end{array}
 
</math>  
 
</math>  
 +
 +
<math>g[m,n]**h[m,n]: \frac{1}{16} X
 +
\begin{array}{cccccccccccc}
 +
0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\
 +
0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\
 +
0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\
 +
1 & 3 & 4 & 4 & 4 & 4 & 4 & 4 & 4 & 3 & 1\\
 +
3 & 9 &12 &12 &12 &12 &12 &12 &12 & 9 & 3\\
 +
4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\
 +
4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\
 +
4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\
 +
4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\
 +
3 & 9 &12 &12 &12 &12 &12 &12 &12 & 9 & 3\\
 +
1 & 3 & 4 & 4 & 4 & 4 & 4 & 4 & 4 & 3 & 1\\
 +
\end{array}</math>
 +
 
----
 
----
  

Latest revision as of 15:07, 2 December 2014


Homework 11, ECE438, Fall 2014, Prof. Boutin


Question 1

Consider the following filter:

$ h[m,n]: \begin{array}{cccc} & m=-1 & m=0 & m=1 \\ n=1&\frac{1}{16} & \frac{2}{16} & \frac{1}{16} \\ n=0&\frac{2}{16} & \frac{4}{16} & \frac{2}{16} \\ n=-1&\frac{1}{16} & \frac{2}{16} & \frac{1}{16} \end{array} $

a) Write a difference equation that can be used to implement this filter.

$ \begin{align} y[m,n] =& \frac{1}{16}(x[m+1,n-1] + 2x[m,n-1] + x[m-1,n-1] \\ & + 2x[m+1,n] + 4x[m,n] + 2x[m-1,n] \\ & + x[m+1,n+1] + 2x[m,n+1] + x[m-1,n+1]) \end{align} $

b) Is this filter separable? Answer yes/no and justify your answer.

Yes. The coefficient matrix of h[m,n] can be written as product of two vectors.

$ \begin{pmatrix} \frac{1}{16} & \frac{2}{16} & \frac{1}{16} \\ \frac{2}{16} & \frac{4}{16} & \frac{2}{16} \\ \frac{1}{16} & \frac{2}{16} & \frac{1}{16} \end{pmatrix} = \begin{pmatrix} \frac{1}{4} \\ \frac{2}{4} \\ \frac{1}{4} \end{pmatrix} \cdot \begin{pmatrix} \frac{1}{4} & \frac{2}{4} & \frac{1}{4} \end{pmatrix} $

Therefore the filter can be decomposed to two 1-D filters.

$ h_1[m] = \frac{1}{4}(\delta[m+1] + 2\delta[m] +\delta[m-1]) $

$ h_2[n] = \frac{1}{4}(\delta[n+1] + 2\delta[n] +\delta[n-1]) $


c) Compute the DSFT H(u,v) of this filter. Sketch the plot of H(u,0). Sketch the plot of H(0,v). What are the characteristics of this filter (low-pass, band-pass, or high-pass)?


$ H_1(\mu) = DTFT\{h_1[m]\} = \frac{1}{4}(e^{j\mu} + 2 + e^{-j\mu}) = \frac{1}{2}(1 + cos\mu) $

$ H_2(\nu) = DTFT\{h_2[n]\} = \frac{1}{4}(e^{j\nu} + 2 + e^{-j\nu}) = \frac{1}{2}(1 + cos\nu) $

Using the separability,

$ H(\mu, \nu) = DSFT\{ h[m,n]\} = H_1(\mu)\cdot H_2(\nu) = \frac{1}{4}(1 + cos\mu)(1 + cos\nu) $

$ H(\mu, 0) = \frac{1}{2}(1 + cos\mu) $

HW11 prob1 1.jpg

So, $ H(\mu, 0) $ is a low-pass filter.


$ H(0, \nu) = \frac{1}{2}(1+cos\nu) $

HW11 prob1 2.jpg

So, $ H(0, \nu) $ is a low-pass filter.


d) What is the output image when this filter is applied to the following image (using symmetric boundary conditions)?

$ g[m,n]: \begin{array}{ccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ \end{array} $

$ g[m,n]**h[m,n]: \frac{1}{16} X \begin{array}{cccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 1 & 3 & 4 & 4 & 4 & 4 & 4 & 4 & 4 & 3 & 1\\ 3 & 9 &12 &12 &12 &12 &12 &12 &12 & 9 & 3\\ 4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\ 4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\ 4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\ 4 &12 &16 &16 &16 &16 &16 &16 &16 &12 & 4\\ 3 & 9 &12 &12 &12 &12 &12 &12 &12 & 9 & 3\\ 1 & 3 & 4 & 4 & 4 & 4 & 4 & 4 & 4 & 3 & 1\\ \end{array} $


Question 2

Consider the following filter:

$ h[m,n]: \begin{array}{cccc} & m=-1 & m=0 & m=1 \\ n=1&-\frac{1}{9} & -\frac{1}{9} & -\frac{1}{9} \\ n=0&-\frac{1}{9} & -\frac{8}{9} & -\frac{1}{9} \\ n=-1&-\frac{1}{9} &- \frac{1}{9} & -\frac{1}{9} \end{array} $

a) Write a difference equation that can be used to implement this filter.

$ \begin{align} y[m,n] =& -\frac{1}{9}(x[m+1,n-1] + x[m,n-1] + x[m-1,n-1] \\ & + x[m+1,n] + 8 x[m,n] + x[m-1,n] \\ & + x[m+1,n+1] + x[m,n+1] + x[m-1,n+1]) \end{align} $


b) Is this filter separable? Answer yes/no and justify your answer.

No. The coefficient matrix of h[m,n] cannot be decomposed to two vectors.


c) Compute the DSFT H(u,v) of this filter. Sketch the plot of H(u,0). Sketch the plot of H(0,v). What are the characteristics of this filter (low-pass, band-pass, or high-pass)?

$ \begin{align} H(\mu, \nu) = DTFT\{h[m,n]\} =& -\frac{1}{9} (e^{j(\mu-\nu)} + e^{j(-\nu)} + e^{j(-\mu-\nu)} \\ & + e^{j(\mu)} + 8 e^{(0)} + e^{j(-\mu)} \\ & + e^{j(\mu+\nu)} + e^{j(\nu)} + e^{j(-\mu+\nu)} ) \end{align} $

$ \begin{align} H(\mu, 0) =& -\frac{1}{9}( e^{j\mu} + e^{-j\nu} \\ & + e^{j\mu} + 8 + e^{-j\nu} \\ & + e^{j\mu} + e^{-j\nu} ) \\ =& -\frac{1}{9}(8+6cos{\mu}) \end{align} $

$ \begin{align} H(0, \nu) =& -\frac{1}{9} (e^{-j\nu} + e^{-j\nu} + e^{-j\nu} \\ & + 8 \\ & + e^{j\nu} + e^{j\nu} + e^{j\nu}) \\ =& -\frac{1}{9}(8+6cos{\mu}) \end{align} $

$ |H(\mu, 0)| = \frac{1}{9}(8+6cos{\mu}) $

HW11 prob2 1.jpg

So, $ H(\mu, 0) $ is a high-pass filter.


$ |H(0, \nu)| = \frac{1}{9}(8+6cos{\nu}) $

HW11 prob2 2.jpg

So, $ H(0, \nu) $ is a high-pass filter.


d) What is the output image when this filter is applied to the following image (using symmetric boundary conditions)?

$ g[m,n]: \begin{array}{ccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ \end{array} $

$ g[m,n]**h[m,n]: \frac{1}{9} X \begin{array}{cccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ -1&-2 &-3 &-3 &-3 &-3 &-3 &-3 &-3 &-2 & 1\\ -2& 5 & 3 & 3 & 3 & 3 & 3 & 3 & 3 & 5 &-2\\ -3& 3 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 3 &-3\\ -3& 3 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 3 &-3\\ -3& 3 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 3 &-3\\ -3& 3 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 3 &-3\\ -2& 5 & 3 & 3 & 3 & 3 & 3 & 3 & 3 & 5 &-2\\ -1&-2 &-3 &-3 &-3 &-3 &-3 &-3 &-3 &-2 & 1\\ \end{array} $



Question 3

Consider the following filter:

$ h[m,n]: \begin{array}{cccc} & m=-1 & m=0 & m=1 \\ n=1&-\frac{1}{8} & \frac{1}{2} & -\frac{1}{8} \\ n=0&-\frac{1}{4} & 1 & -\frac{1}{4} \\ n=-1&-\frac{1}{8} & \frac{1}{2} & -\frac{1}{8} \end{array} $

a) Write a difference equation that can be used to implement this filter.

$ \begin{align} y[m,n] =& -\frac{1}{8}x[m+1,n-1] + \frac{1}{2}x[m,n-1] - \frac{1}{8}x[m-1,n-1] \\ & -\frac{1}{4}x[m+1,n] + x[m,n] -\frac{1}{4}x[m-1,n] \\ & -\frac{1}{8}x[m+1,n+1] + \frac{1}{2}x[m,n+1] -\frac{1}{8}x[m-1,n+1] \end{align} $

b) Is this filter separable? Answer yes/no and justify your answer.

Yes. The coefficient matrix of h[m,n] can be written as product of two vectors.

$ \begin{pmatrix} -\frac{1}{8} & \frac{1}{2} & -\frac{1}{8} \\ -\frac{1}{4} & 1 & -\frac{1}{4} \\ -\frac{1}{8} & \frac{1}{2} & -\frac{1}{8} \end{pmatrix} = \begin{pmatrix} \frac{1}{2} \\ 1 \\ \frac{1}{2} \end{pmatrix} \cdot \begin{pmatrix} -\frac{1}{4} & 1 & -\frac{1}{4} \end{pmatrix} $

Therefore the filter can be separated into two 1-D filters.

$ h_1[m] = -\frac{1}{4}\delta[m+1] + \delta[m] -\frac{1}{4}\delta[m-1] $

$ h_2[n] = \frac{1}{2}\delta[n+1] + \delta[n] +\frac{1}{2}\delta[n-1] $

c) Compute the DSFT H(u,v) of this filter. Sketch the plot of H(u,0). Sketch the plot of H(0,v). What are the characteristics of this filter (low-pass, band-pass, or high-pass)?


$ H_1(\mu) = DTFT\{h_1[m]\} = -\frac{1}{4}e^{-j\mu(-1)} + e^{-j\mu(0)} -\frac{1}{4}e^{-j\mu(1)} = 1-\frac{1}{2}cos\mu $

$ H_2(\nu) = DTFT\{h_2[n]\} = \frac{1}{2}e^{-j\nu(-1)} + e^{-j\nu(0)} +\frac{1}{2}e^{-j\nu(1)} = 1+cos\nu $

Using the separability,

$ H(\mu, \nu) = DSFT\{ h[m,n]\} = H_1(\mu)\cdot H_2(\nu) = (1-\frac{1}{2}cos\mu)(1+cos\nu) $

$ H(\mu, 0) = 2(1-\frac{1}{2}cos\mu) $

HW11 prob3 1.jpg

So, $ H(\mu, 0) $ is a high-pass filter.


$ H(0, \nu) = \frac{1}{2}(1+cos\nu) $

HW11 prob3 2.jpg

So, $ H(0, \nu) $ is a low-pass filter.


d) What is the output image when this filter is applied to the following image (using symmetric boundary conditions)?

$ g[m,n]: \begin{array}{ccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 1 & 1 & 1 & 1 & 1 & 0 & 0 & 0\\ 0 & 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ \end{array} $


$ g[m,n]**h[m,n]: \frac{1}{9} X \begin{array}{cccccccccccc} 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 &-1 & 3 & 2 & 2 & 2 & 3 &-1 & 0 & 0\\ 0 &-1 & 1 & 8 & 6 & 6 & 6 & 8 & 1 &-1 & 0\\ -1& 1 & 7 & 9 & 8 & 8 & 8 & 9 & 7 & 1 &-1\\ -3& 8 & 9 & 8 & 8 & 8 & 8 & 8 & 9 & 8 &-3\\ -4&12 & 8 & 8 & 8 & 8 & 8 & 8 & 8 &12 &-4\\ -4&12 & 8 & 8 & 8 & 8 & 8 & 8 & 8 &12 &-4\\ -4&12 & 8 & 8 & 8 & 8 & 8 & 8 & 8 &12 &-4\\ -3& 9 & 6 & 6 & 6 & 6 & 6 & 6 & 6 & 9 &-3\\ -1& 3 & 2 & 2 & 2 & 2 & 2 & 2 & 2 & 3 &-1\\ \end{array} $




Discussion

You may discuss the homework below.

  • write comment/question here
    • answer will go here

Back to ECE438, Fall 2014, Prof. Boutin

Alumni Liaison

Ph.D. on Applied Mathematics in Aug 2007. Involved on applications of image super-resolution to electron microscopy

Francisco Blanco-Silva