You are on page 1of 2

Deblurring

Adam Leach
March 25, 2014

1
Let L(, ) be the original luminance
Let H(, , , ) be the weighting function of our antenna
Let A(, ) be our measured luminance
Then:

A(, ) =

L(, ) H(, , , ) d d
0

We can decompose our spherical functions into the real spherical harmonics
H(, , , ) =

X
l
X

Hlm (, ) Ylm (, )

l=0 m=l

L(, ) =

X
l
X

Llm Ylm (, )

l=0 m=l

Thus:
Z

A(, ) =
0

2 X

l
X

Llm Ylm (, )

l=0 m=l

X
l
X

Hlm (, ) Ylm (, ) d d

l=0 m=l

Due to the orthonormality of the real spherical harmonics, and the linearity of
the integral function:
A(, ) =

X
l
X

Llm Hlm (, )

l=0 m=l

If instead of using the infinite sum spherical harmonic decomposition, we use a


finite approximation up to k terms, we are left with:
A(, ) =

k X
l
X
l=0 m=l

Llm Hlm (, )

As we know A(, ) and Hlm (, ), we then have a system of n linear equations


from which to determine Llm . Of course, our values for A(, ) will not be
perfect, so we shall need to introduce variables 0 ...n for the n samples from
A(, ) and Hlm (, ) in order to determine Llm via an optimisation problem
minimising
02 + 12 + ... + n2
Once Llm has been found for all values for l and m, all one needs to do to
construct an image of the original luminance function is to then compute
L(, ) =

X
l
X
l=0 m=l

Llm Ylm (, )

You might also like