[daip] rmsd

[Leonia Kogan]

Lincoln, Lynn,

Lincoln wrote:
|The description of how RMS is measured (via iterations) sounds different
|from how it is done via histograms in IMEAN or other routines?

Both algorithms should require using the box  surrounding the given 
pixel to estimate RMS.
Both algorithms allow to circumwent a possible big feature at the box.
Both algorithm may require a lot of computing time:
RMSD -many(actualy several) iterations
IMEAN- fitting two dimensional Gausian using non linear least square, 
which can require several iterations also
                with more calculations at eac iterations (inverse matrix 
...)

So I am not sure that IMEAN's algorithm  will be faster at RMSD when  it 
is asked to repeat it  a lot of time (for each pixelo of image say 
1024*1024.

Using XINC, YINC (see the explain at my previous message) reduces the 
computational time very much

Regards

Leonia




Leonia Kogan wrote:

> Lincoln, Lynn,
>
> The following is my quick answers on the questions, inserted into the 
> Lincoln's text.
>
> Lincoln Greenhill wrote:
>
>> Hi Leonia,
>>
>> I am sorry that I missed you when I called on Monday to talk about RMSD.
>> I'm on my way to Tucson at the moment.  I will try to reach you again
>> Thursday or Friday when there are breaks at the meeting.
>>
>> If Lynn prefers, she could contact you instead of me.  If its by email,
>> please keep me on the cc-list.
>>
>> The basic questions are:
>>
>> The description of how RMS is measured (via iterations) sounds different
>> from how it is done via histograms in IMEAN or other routines?
>>  
>>
> First,  rms for each pixel is calculated based on the points incide of 
> the box centered at the given pixel and sized  by
> IMSIZE.
> Second,  at the first iteration mean and rms are calculated using all 
> points inside of the box. If the pixel value differs
> from the mean more than 3*rms then such a pixel is excluded from the 
> considerations at the following itteration.
> Third,  The iteration process is ended when none point is excluded
>
>> How are XINC and YINC used?
>>  
>>
> RMSD requires a lot of computing time. XINC, YINC are used to make 
> this time less
> The rms is calculated using multii iteration process only for the 
> pixels following with steps XINC, YINC at the
> relevant exes. For all other pixels rms is calculated using 
> interpolation.
>
>> In terms of potential upgrades to the software that were  not
>> implemented when it was first assembled, is
>> there room in an obvious way to improve the runtime efficiency of the
>> program?
>>  
>>
> I'll look. But I doubt
>
>>
>> Best,
>> Lincoln
>>
>>
>>  
>>
>