[R-sig-Geo] Neighborhood filtering

[Roland Löwe]

Nice, thanks. That's what I'm looking for. 
It looks, like the for-loops may run faster (I tested on 10000 iterations of the above operation) but I'll try again  with the actual data.

------------------------------

Message: 11
Date: Tue, 23 Aug 2011 10:41:43 +0200
From: Mathieu Rajerison <mathieu.rajerison at gmail.com>
To: r-sig-geo at r-project.org
Subject: Re: [R-sig-Geo] Neighborhood filtering
Message-ID:
	<CAGfc75kq3-LZt1GvPcQRi8OjD_2t1ok9ECEkCWWSkqWTgPZQzA at mail.gmail.com>
Content-Type: text/plain

Hi,

I think all you need is in raster package.

You could transform your matrix into a RasterLayer object

Then apply focalFilter function on your RasterLayer using your matrix of weights

http://rgm2.lab.nig.ac.jp/RGM2/func.php?rd_id=raster:focalFilter

Hope this helps

2011/8/23 Roland L?we <rolo at imm.dtu.dk>

> This may be a classic problem but I couldn't find a solution in the 
> lists so far. I'm trying to run a filter on radar data, where a pixels 
> result value is predicted as weighted average of the values in a 3x3 
> neighborhood in the current time step.
> For one time step the operation could look like
>
> filter<-matrix(rep(1/9,9),3)
>          [,1]      [,2]      [,3]
> [1,] 0.1111111 0.1111111 0.1111111
> [2,] 0.1111111 0.1111111 0.1111111
> [3,] 0.1111111 0.1111111 0.1111111
>
> base<-matrix(runif(100,0.0,2),10)
>      [,1]       [,2]      [,3]       [,4]       [,5]      [,6]       [,7]
>     [,8]       [,9]      [,10]
>  [1,] 1.2748495 1.64100480 1.7629353 1.04471073 0.17636041 0.7920916
> 1.81948900 0.28150594 0.40615944 0.33799927  [2,] 0.9134852 1.04947132 
> 1.7847194 0.28143589 0.25744083 1.4310364
> 1.48369795 1.04006404 0.51538211 1.54781491  [3,] 1.4310985 1.01676632 
> 1.8475540 1.97060630 0.01470109 0.2014272
> 1.79969115 1.04854042 0.90953386 0.02998074  [4,] 1.7304672 0.93514790 
> 0.1203939 0.69291092 0.34618531 1.5409637
> 0.58926745 1.61994100 1.66063354 0.69013138  [5,] 0.7535980 0.64312043 
> 1.2213976 0.03293221 1.70696788 1.6062746
> 0.06455516 0.62512456 1.94994886 0.52073311  [6,] 1.6559531 0.35279738 
> 0.5878466 1.99312250 1.98113265 0.4076653
> 1.60588111 1.77600453 0.73479768 1.99889307  [7,] 0.1142795 0.22586611 
> 0.2603767 0.33145144 1.37324328 0.8435238
> 0.78651141 0.46745984 0.93032790 1.20699560  [8,] 1.9036901 1.71863581 
> 0.6056311 1.83853173 1.82158619 0.3230073
> 0.17650283 0.02147692 0.02384052 1.56027144  [9,] 1.8832814 1.38824498 
> 0.7263898 1.63432910 1.65317872 1.4431693
> 1.06139030 1.08942924 0.15570795 1.21371980 [10,] 0.5374888 0.03651226 
> 1.3404380 1.23004404 1.99562437 1.4025763
> 1.35752389 0.39377822 1.03457161 0.90385665
>
> result<-matrix(ncol=8,nrow=8)
> for (i in c(2:9)){
>  for (j in c(2:9)){
>  result[(i-1),(j-1)]<-sum(base[(i-1):(i+1),(j-1):(j+1)]*filter)
> }
> }
>
>          [,1]      [,2]      [,3]      [,4]      [,5]      [,6]      [,7]
>    [,8]
> [1,] 1.4135427 1.3776893 1.0156071 0.6855345 0.8862151 1.0997271 
> 1.0337849
> 0.6796645
> [2,] 1.2032338 1.0776673 0.8128831 0.7485231 0.8516012 1.1949588 
> 1.1851946
> 1.0068913
> [3,] 1.0777271 0.9423144 0.8837388 0.9014410 0.8744482 1.0106428 
> 1.1408040
> 1.0060631
> [4,] 0.8889691 0.7310744 0.9647655 1.1453506 1.0943215 1.0928531 
> 1.1806838
> 1.2862453
> [5,] 0.6461373 0.6276568 1.0542745 1.1418126 1.1528617 0.9092223 
> 0.9934012
> 1.1344761
> [6,] 0.8250085 0.8793621 1.1992136 1.2125849 1.0354504 0.7120037 
> 0.7247559
> 0.9688964
> [7,] 0.9807106 0.9699396 1.1383020 1.2513357 1.0535681 0.6902746 
> 0.5236274
> 0.7410255
> [8,] 1.1267014 1.1687508 1.4273059 1.4824497 1.2482844 0.8076505 
> 0.5904691
> 0.7107391
>
>
> I will have to run this operation about 300.000 times in a parameter 
> estimation routine, so I was wondering if anyone could point me at an 
> efficient way of filtering that avoids the nested for loop.
> Thanks.
> Roland
>
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

	[[alternative HTML version deleted]]



------------------------------

_______________________________________________
R-sig-Geo mailing list
R-sig-Geo at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-geo


End of R-sig-Geo Digest, Vol 96, Issue 21