For which is zero mean and uncorrected with and are zero and an estimate of is obtained as

	(6.3.5)

where obtained from known or measured properties of

From above estimate is not guaranteed to be non-negative. To ensure this, is assured zero if is

Given an estimate of there are many different ways to estimate .

One method that is frequently used and is consistent with the idea that mean square error filters such as Wiener filters are zero phase is to approximate by .

So that,

and

In equation (6.3.2), can be viewed as a periodgram and since variance of periodgram is known to be very large subtraction of does not reduce noise sufficiently. To reduce more noise at the expense of more signal distortion, with is often subtracted from In this case, the estimate is obtained from

	(6.3.6)

is a parameter that controls the level of noise reduction.

Since equation (6.3.6) is used for each subimage and the processed subimages are combined to form the whole image, the method is called "short space" spectral subtraction. This method can be viewed as enhancing the SNR. Since the same function Is subtracted from in both high detail and low detail region, the subtraction has a relatively small effect in high detail/region, where is relatively large. In low detail regions where is relatively small & consists primarily of the noise component, the subtraction eliminates a relatively large portion of .

Separable overlapped triangular window is used in this method.

Phase retrieval Method

The algorithm is shown as a flow chart. This also has been observed to converge to the desired solution when the initial estimate used is quite accurate or the signal has a special characteristic such as triangular region of support.

The magnitude only reconstruction problem specifies within a sign factor, translation and rotation by . Hence more than one solution is possible .Imposing an initial estimate sufficiently close to a possible solution or imposing additional constraints such as triangular region of support appear to prevent the iterative process from wandering.