bilinear interpolation of 2D matrix

I am trying to apply Bilinear/bicubic interpolation on my data set which is in text files. These text files are in a folder and name output_00.text to output_23.text. Each text file consist on three columns. First is Latitude, second is longitude and third column is temperature value at this latitude and longitude(position over earth).

Temperature column contain -9999.000 as not a number or NaN values. This NaN value appear randomly in each file. Some times it appear over start, some times in middle of data, and some times at the end in the file. It is random in appearance.

I want to interpolate these NaN values with bilinear/ bicubic interpolation technique.

This code will read each text file and interpolate it with bilinear method and save it with method_00.text.

My text files has been attached

On google i found a tool for bilinear interpolation over a image or matrix. For the time shake this code can be modify to my requirement. But How? This code as been attached with this question. Link of this tool is here http://www.mathworks.com/matlabcentral/fileexchange/43533-bilinear-interpolation-of-an-image-or-matrix

Thanks always for this kind assistance

[Stephen23]

@ Muhammad Usman Saleem: you have already asked this question:

http://www.mathworks.com/matlabcentral/answers/265845-interpolations-nearest-neighbor-and-bilinear-interpolation-from-multiple-text-files

And you have already been told that Because of your missing data you do not have gridded data, but scattered data points, so therefore you can only use a scattered interpolant. You cannot use bilinear/ bicubic interpolation.

Now you asked this question again, and the two answers below have told you exactly the same thing as I did: that you need to use a scattered interpolant. You got the same answer here too.

It seems that you misunderstand what bilinear/ bicubic interpolation can do: it cannot interpolate your data with missing values. It does not matter how many times you ask this question, the problem is your wrong understanding of those interpolation methods. If you spent some time trying to understand how bilinear/ bicubic interpolation work then you would understand why you cannot use them with your data. This would be a good place to start:

https://en.wikipedia.org/wiki/Bilinear_interpolation

[Muhammad Usman Saleem]

@Stephon thanks for your kind contribution always.This question was answered with interpolat1 but now i am asking about interpolation in 2d. Consist on bilinear or scatter.

[Stephen23]

The OP continues this topic here:

http://www.mathworks.com/matlabcentral/answers/275765-rms-error-and-mean-absolute-error-from-text-files

Answer 1

[Mike Garrity]

Here's one approach.

Create an XY grid, but put nans in some of the Z values.

[x,y] = meshgrid(linspace(-3,3,40));
z = peaks(x,y);
m = unique(randi(numel(z),[1 50]));
z(m) = nan;
surf(x,y,z)

Now use scatteredInterpolant to fill in the missing values.

m = isnan(z(:));
F = scatteredInterpolant(x(~m),y(~m),z(~m),'linear');

z2 = z;
z2(m) = F(x(m),y(m));
surf(x,y,z2)

What's going on there is that I created a scatteredInterpolant from the locations where there weren't nans (i.e. x(~m),y(~m),z(~m)), and then I applied it at the locations where there are nans (i.e. x(m),y(m)). Does that make sense?

[Muhammad Usman Saleem]

yes this makes sense. But please note that my question is about bilinear and bicubic interpolation method. Scatteredinterpolant only carry linear, natural and nearest interpolation i want to apply bilinear/ bicubic interpolation over my data set. Thanks for your kind assistance. Please modify this code to bilinear and bicubic interpolation method

[Stephen23]

@Muhammad Usman Saleem

It does not matter how many times you ask, you cannot get the impossible. Bilinear/ bicubic interpolation requires all data on a grid. You have missing data. Ergo you cannot use these methods. This has been explained to you multiple times.

[Muhammad Usman Saleem]

thanks Stephon i know i am absolutely wrong every time.

Answer 2

[Image Analyst]

If your data are in a rectangular grid (i.e., an image), you can use imresize() (in the Image Processing Toolbox) to do bilinear interpolation, or a variety of other interpolations. If they're not in a grid, use scatteredInterpolant like Mike showed you.

[Muhammad Usman Saleem]

@Image please check my data set. I have attached it to this question. I have used scatteredinterpolant for this data sets. But as this holds only linear, natural and nearest interpolation method not bilinear, bicubic interpolation method. On google i find the code which i have attached with this post. Do not know how can i use this code for my requirements?

[Image Analyst]

Take the x and y numbers and multiply by 10 to get pixels. Then assign the 3rd column as image values. Then use imresize().

[Muhammad Usman Saleem]

totally unable to get your answer?

Is linear interpolation in scatteredinterpolant is bilinear interpolation?? What about bicubic interpolation in scatteredinterpolant?

[Image Analyst]

See attached.

[Muhammad Usman Saleem]

thanks for you reply @image.

But this will be modification to my shared code. I want to modify this code so that i can read all test files and interpolation through bilinear and bicubic interpolation and then save it. This attachment only producing my text file as image. And i do not know what will be input in my code