[Developers] common offset prestack processing

[Kristofer Tingdahl]

Dear fellow developers,
>
> as opendtect begins to handle prestack data, I am starting to 
> speculate about porting some algorithms into pre-stack domain.
>
> Some of those would require common-offset sorting of input: each 
> common offset cube would ideally be processed separately and all 
> offsets outputs collected together would yield the final resulting CDP.
>
> I have not inspected the 3.2.1 code closely enough, but I noticed that 
> there is a new attribute that has a prestack input and computes avo 
> attributes (among other things): such attribute has pre-stack input 
> but post-stack output, so it does not prove common-offset processing 
> is possible yet.
>
> What is actually possible in 3.2.1 and what will the future bring in 
> that respect?
>
Yes, we now have pre-stack, but the design of OpendTect is all around 
interpretation. Thus, we made a model of what Pre-stack data 'is', 
around the simple: 'we need to access it for interpretation' idea. Give 
a position (BinID in 3D, line name and trace number in 2D), and you'll 
get the CMP gather back.

We'd expect the pre-stack data in the store to be migrated and 
everything: basically, the data that when stacked will deliver the 
post-stack cube. That is the only really usable data in interpretation.

What you're describing is actually processing, and since you are forced 
to write ps-data in inl/crl/offset sequence, you cannot do what you 
intend to do directly. To be 100% correct, you can write inlines in 
random order, as long as you write the individual inlines in crl/offset 
sorting. So, if your algorithm can do one inline at a time, and you have 
memory enough to buffer an inline, you can do that. The pre-stack reader 
can be used for random read without limitations, as described in the 
initial paragraph. If you cannot do one inline at a time, you can make a 
batch-type program where you use the prestack reader to read in whatever 
sorting you want, and write out cbvs volumes, one per offset. Then, you 
merge them in a final step and remove the intermediate volumes.

Best regards,


Kristofer Tingdahl

-- 
Dr. Kristofer Tingdahl
Executive Vice President - Americas
dGB Earth Sciences
1 Sugar Creek Center BLVD #935; Sugar Land, TX 77478; USA
+1 281 240 3939