Levenberg-Marquardt Algorithm

tejaswi prakash tejaswidprakash at gmail.com
Wed Apr 11 07:09:03 EDT 2018


I am sorry, but I thought Levenberg marquardt was used quite  bit in Image
registration. Computing/refining homographies between two related views for
instance.

On Wed, Apr 11, 2018 at 12:49 PM, Christian Gollwitzer <auriocus at gmx.de>
wrote:

> Am 11.04.18 um 08:38 schrieb Priya Singh:
>
>> I have two 2D arrays one R and another T (which is also a 2D array).
>> Do you know how can I fit T with R in order to find central
>> coordinate x0,y0 for T relative to R???
>>
>> So the main question is do you know in python how can I fit two 2D arrays
>> to find
>> x0,y0 for one array relative to other. I shall use LM fit in python. But
>> for fitting, I need to have some fittable model but here I am having only
>> two 2D arrays. I know simple cross-correlation would have solved my problem
>> but I have been instructed to do fitting using one array to other.
>>
>
>
> The request is nonsense. LM fits an analytical model to data, if you don't
> have an analytical model, you need another tool. Cross correlation is
> widely used and works well for many such tasks.
>
> In principle you could also interpolate the one array to new coordinates,
> e.g. using scipy.ndimage.interpolation.shift, and minimize the sum of
> squared differences. But still LM is the wrong tool here, it would get
> trapped in local minima soon, and it uses derivatives. Look for "image
> registration" to find typical algorithms used in this context.
>
>
>
>         Christian
>
> --
> https://mail.python.org/mailman/listinfo/python-list
>


More information about the Python-list mailing list