Using sgolayfilt in derivative analysis of absorption images

I want to apply this equation to my absorption images at 655 nm

This code enables me to do that

fileDirectory = "/Volumes/Extreme_SSD/HPLC validation images";
filename = dir(fullfile(fileDirectory, '*aph.nc'));
nFileDirectory = length(filename);

%create directories for the chlorophyll (yearwise)
outputFolder = "/Volumes/Extreme_SSD/HPLC validation images";

for i=1:nFileDirectory
    nFileAph = strcat(filename(i).folder, "/", filename(i).name);
    ncdisp(nFileAph);
    info = ncinfo(nFileAph);
    dim = cell2mat({info.Variables.Dimensions});
    ncol = dim(2).Length;
    nrow = dim(1).Length;
    aph654 = ncread(nFileAph, "a_ph_654");
    aph656 = ncread(nFileAph, 'a_ph_656'); 
    
    deltaLambda = 1;  % nm

% derivative of 655
    aph_deriv_655 = (aph656 - aph654) / (2 * deltaLambda);
end

now I am confused how to use filter such as sgolayfilt (window = 7, wavlength gap 5 nm) in the derived image.

[Walter Roberson]

dim = cell2mat({info.Variables.Dimensions});

It seems likely to me that the above line could be coded as

dim = [info.Variables.Dimensions];

Answer 1

[Star Strider]

The sgolayfilt function is used in signal processing of one-dimensional signals, not image processing of two-dimensional arrays. It creates a FIR smoothing filter (essentially a lowpass filter), however it is not obvious to me how your 'window' and wavelength gap would translate to framelength and polynomial order. It operates on each column of a matrix. (As a general rule, I use a third-order polynomial, and experiment with the framelength to get the result I want.) Filtering your matrix using sgolayfilt, transposing it and filtering it a second time using the same sgolayfilt call, and transposing it back to the original orientation could work. You will have to experiment to see if that produces an acceptable result.

The imfilter function can do image filtering, however it requires a multidimensional filter, and sgolay creates a unidimensional filter. It is not obvious how vector operations on two sgolay filters (not using sgolayfilt) would work to create a two-dimensional filter and retain the desired spectral characteristics.

What do you actually want to do?

[Arnab Paul]

smoothening the images. looks very noisy. as I have field observation of spectra (1D) that smoothened using sgolayfilt so I am trying to use sgolayfilt for the images as well.

[Star Strider]

I do not have an example image to experiment with.

I suggest that you do the bi-directional filtering as I suggested. Make appropriate changes to the sgolayfilt arguments to get the reswult you want.

Another option is to use imfilt with an fspecial filter to remove the noise. A Gaussian filter should work.

[Arnab Paul]

is their any example how to filter using Gaussian filter?

[Star Strider]

Not directly. See the documentation on imfilter for an example using a 'motion' filter.

It would probably be best for you to use the imgaussfilt function, since the Gaussian fspecial filter is no longer recommended.

(Answers is having serious problems currently. My apologies for the delay.)