smoothn

Smooths N-D input data using 'box' or 'gaussian' filter using a specified convolution kernel.

Description

SMOOTHN Smooth N-D data
   Y = SMOOTHN(X, SIZE) smooths input data X. The smoothed data is
       retuirned in Y. SIZE sets the size of the convolution kernel
       such that LENGTH(SIZE) = NDIMS(X)

   Y = SMOOTHN(X, SIZE, FILTER) Filter can be 'gaussian' or 'box' (default)
       and determines the convolution kernel.

  Y = SMOOTHN(X, SIZE, FILTER, STD) STD is a vector of standard deviations
       one for each dimension, when filter is 'gaussian' (default is 0.65)

See also SMOOTH3

(It is an extension of Mathworks' SMOOTH3 from 3 to N dimensions)

• Acknowledgements: This submission has inspired the following:; fast running mean
• MATLAB release: MATLAB 5.3.1 (R11.1)

[Carlos Vargas]

 

[glacier cao]

Good!

[Zoltan Nadasdy]

 

[Ed Chao]

Works like it should and verified to be accurate. Matches the IDL function "SMOOTH". But this function is *slow*. A 1 sec. IDL program takes about 20 seconds with this function.

[Psul Griffin]

slow, but good results.

[Kent Conover]

Failed to finish a smooth of a 1000 point vector when run on Mac OS X system with Matlab Version 7.1.0.183 (R14) Service Pack 3.

[Fco. Camacho]

Good. (For the box filter (enough for me) in 1 and 2d i prefer moving_average by Carlos Vargas, it's a ferrari)

[y c]

GOOD, but gives problems when plotting a 0°-360° longitude contour map.
0° longitude data do not coincide with 360° longitude data.there is a shift of the isolines.