Solved

# 2D convolution

Posted on 2004-11-06
Medium Priority
628 Views
Hey experts!

x(n1, n2) = 1      for n1&#8805;0, n2&#8805;n1
0             otherwise

I want to find the colvolution of x with itself (x is a 2D signal)
0
Question by:mte01
• 7
• 7
• 3
• +1

LVL 3

Author Comment

ID: 12513166
"and &#8805;" would mean greater than or equal
0

LVL 3

Author Comment

ID: 12513171
Here is the question written properly:

(n1, n2) = 1     for n1>=0, n2>=n1
0             otherwise

I want to find the colvolution of x with itself (x is a 2D signal)
0

LVL 31

Expert Comment

ID: 12513486
mte01,
This function is 1 in the upper right quadrant and zero else where, ie non-zero over an infnite  domain, the overlap region will always be infnite or empty so giving an infinite  convolution integral. Is there some other assumption that I am not aware of with signal convolutions?  I will look again later today, a bit busy now.
0

LVL 31

Expert Comment

ID: 12513539
I mean upper right octant (ie a triangular region )
0

LVL 3

Author Comment

ID: 12513736
Yes the result of the convolution integral will be also infinite, but it can be defined in a summation form or something like that.....
0

LVL 27

Expert Comment

ID: 12514260
I think that if you want the convolution to be a series of terms, you must express the function itself in a series of terms
0

LVL 27

Expert Comment

ID: 12514272
I may not be visulizing this correctly, but in this case will not the self-convolution just be itself?
0

LVL 3

Author Comment

ID: 12514286
you can say that x = sum(from n1=0 to inf) sum(from n2=n1 to inf) (1)

no the convolution of x with itself won't be x.........
0

LVL 31

Expert Comment

ID: 12514930
With convolution aren't you reversing one of the functions and then comparing the two?  If the two functions being compared are identical then you are comparing two identical signals, which is the auto-correlation of the two?  In the frequency domain you will get a spike at 0, followed by spikes at specific frequencies if there is any periodicity in the signal.  (It's years since I studied this so apologies for my naivety).
0

LVL 3

Author Comment

ID: 12514955
Yes right, and what I want to know are these fequencies at which the "spikes" occur, that is, the values at which the convolution rseult will be 1 (in a summation form)
0

LVL 31

Expert Comment

ID: 12515079
I can't picture the shape you have there.  If it is a triangle then you will have exponentially decaying spikes, the first one at the origin which is of infinite size (because you've stepped from 0 to 1 infinitely quickly).

You should be able to plug the figures into a Fourier Transform of your function to get the sizes of the rest of the spikes.  It will have e in it somewhere!
0

LVL 31

Expert Comment

ID: 12515093
...exponentially decaying positive and negative spikes....
0

LVL 3

Author Comment

ID: 12515094
The Fourier Transform of this function does not exist, so we have to do a convolution in the classical addition-multiplication way, or we should try the 2D Z-transform of this function (I am right now thinking of doing that method)
0

LVL 31

Expert Comment

ID: 12515442
This is what I reckon    (using ** for convlolution operator)

1     for n1>=0, n2>=n1
x(n1, n2) =
0     otherwise

The convoultion of x(n1, n2) with its self is

x(n1, n2)**x(n1, n2) = sum(l=0,inf)sum(m=0,inf){ x(l,m)*x(l-n1,m-n2) }

The non-zero region of x(n1, n2) is the upper right hand triangle of the plane,

The non-zero region of the function x(-n1,-n2) which is the the reflection of this region about both axises (ie lower left central traingle).

For for n1>=0, n2>=n1 the sumation x(n1, n2)* x(n1, n2) is over the intersection the lower triangle shifted up to the point  (n1, n2).

This region is a parallelogram with its lower point at the origin and upper point at (n1,n2) of height n2-n1.

By my estimation the number of values in this region is n1*(n2-n1) each of value 1 so giving

n1*(n2-n1)     for n1>=0, n2>=n1
x(n1, n2)** x(n1, n2)  =
0               otherwise
0

LVL 31

Expert Comment

ID: 12515445
shld read the sumation x(n1, n2)**x(n1, n2)
0

LVL 31

Accepted Solution

GwynforWeb earned 2000 total points
ID: 12515471
A couple of errors above

Here are  errors corrected,

1     for n1>=0, n2>=n1
x(n1, n2) =
0     otherwise

The convoultion of x(n1, n2) with its self is

x(n1, n2)* x(n1, n2) = sum(l=0,inf)sum(m=0,inf){ x(l,m)x(l-n1,m-n2) }

The non-zero region of x(n1, n2) is the upper right hand triangle of the plane,

The non-zero region of the function x(-n1,-n2) which is the the reflection of this region about both axises (ie lower left central triangle).

For for n1>=0, n2>=n1 the sumation x(n1, n2)**x(n1, n2) is over the intersection the lower triangle shifted up to the point  (n1, n2).

This region is a parallelogram with its lower point at the origin and upper point at (n1,n2) of height n2-n1.

By my estimation the number of values in this region is (n1+1)(n2-n1+1) each of value 1 so giving

(n1+1)(n2-n1+1)     for n1>=0, n2>=n1
x(n1, n2)* x(n1, n2) =
0                 otherwise

0

LVL 31

Expert Comment

ID: 12515836
x(n1, n2)* x(n1, n2)    shld be    x(n1, n2)**x(n1, n2)  ie the convolution
0

LVL 3

Author Comment

ID: 12516683
Amazing!!!!!......Excellant!!!!!!...this is exactly what I wanted.......
0

LVL 31

Expert Comment

ID: 12517685
Thanks for the points, it's a suprising result. For the archival purposes I have a further typo, ie shld read  x(n1, n2)**x(n1, n2) = sum(l=0,inf)sum(m=0,inf){ x(l,m)x(n1-l,n2-m),  the analysis is correct though and based on the correct formula.
0

## Featured Post

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Introduction On a scale of 1 to 10, how would you rate our Product? Many of us have answered that question time and time again. But only a few of us have had the pleasure of receiving a stack of the filled out surveys and being asked to do somethiâ€¦
Aerodynamic noise is the cause of the majority of the noise produced by helicopters. The inordinate amount of noise helicopters produce is a major problem in the both a military and civilian setting. To remedy this problem the use of an aerogel coatâ€¦
This is a video describing the growing solar energy use in Utah. This is a topic that greatly interests me and so I decided to produce a video about it.
Although Jacob Bernoulli (1654-1705) has been credited as the creator of "Binomial Distribution Table", Gottfried Leibniz (1646-1716) did his dissertation on the subject in 1666; Leibniz you may recall is the co-inventor of "Calculus" and beat Isaacâ€¦
###### Suggested Courses
Course of the Month16 days, 4 hours left to enroll