[OpendTect_Users] Seismic Attribute Normalization (mohsen seyed ali)

[Julien Moreau]

Dear Mohsen,

I had to do such manipulation and I discovered that it is not really  
straightforward to blend attributes and interpret them as probabilities.

First of all, changing the repartition of your data might be  
mathematically wrong. Not all the distributions can be fit to a gauss  
curve. The best example being your amplitude histogram if it contains  
extreme amplitudes which are your target. Applying gauss on it will  
make these anomalous amplitudes insignificant.

The second point is the problem of rocks themselves, it is a discrete  
data. Often, we search to image if it is this rock or that other (or a  
fault etc). in this case too, you want to have a sharp distinction  
between the data not interesting and your target.

There are several geostatistic solutions to this problem but it is not  
really the purpose of OpendTect (and I find it often rather complex).

So to come to the point, what I did was maybe a bit wrong  
statistically but relatively straightforward:
   - visualise the attributes via RGB blending. It gave me the  
polarity of the attribute, meaning I knew which extreme of the data  
range my target laid. You can use the colour charts for that or  
multiple cross-plots. If your target is in the central part of your  
data span you can try to apply an inverse function. If it is the two  
ends of the dataset, RMS or abs will do too. NB: RMS will "gauss" a  
bit your data.
   - According to this observation, set an attribute probability  
ranging from 0 to 1. 1 being the target maximum probability. For that  
use the functions max or min and 'standardise' your data. X =  
X/max(X); max(X) being your target.
   - multiply all the probabilities and you will get a general  
probability. You can also use harmonic means and other more smooth  
ways. However, I suggest to start with a regular multiplication.
   - if you are not satisfied, use coefficients but maintain the  
individual probabilities between 0 and 1. The coefficients can be  
logs, powers, and constants, it depends if you want or not to spread  
your data toward 0 or 1.

All this can be set relatively quickly with the attribute mathematics .

Hope it helps,

Best wishes,

Julien

Quoting users-request at opendtect.org:

> Send Users mailing list submissions to
> 	users at opendtect.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> 	http://lists.opendtect.org/mailman/listinfo/users
> or, via email, send a message with subject or body 'help' to
> 	users-request at opendtect.org
>
> You can reach the person managing the list at
> 	users-owner at opendtect.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Users digest..."
>
>
> Today's Topics:
>
>    1. Seismic Attribute Normalization (mohsen seyed ali)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sat, 20 Aug 2011 16:25:59 +0430
> From: mohsen seyed ali <mohsen.seyedali at gmail.com>
> Subject: [OpendTect_Users] Seismic Attribute Normalization
> To: users at opendtect.org
> Message-ID:
> 	<CAJ8ygSsCpafHhpNd2K-5Fhg-vEbtmQ9r0Ju3khiQrpAsJWRZAQ at mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Hello everybody,
>
> I am going to cluster (un-supervised) some seismic attributes which are
> calculate by opendtect. The calculated attributes have different value
> ranges which causes the  multi-attribute clustering encounters some
> problems. To solve this problem, it seems all the involved attributes must
> be normalized in an specified range. Is there any way in opendtect to
> normalize the seismic attributes in an specified range? (such as gaussian
> distribution)
>
> Thanks all,
> Mohsen.
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:  
> http://lists.opendtect.org/pipermail/users/attachments/20110820/bc60e3ee/attachment.html
>
> ------------------------------
>
> _______________________________________________
> Users mailing list
> Users at opendtect.org
> http://lists.opendtect.org/mailman/listinfo/users
>
>
> End of Users Digest, Vol 71, Issue 5
> ************************************
>
>



-- 
Dr Julien Moreau
Postdoctoral Researcher
Institute of Seismology
University of Helsinki