Matlab

Reading PAM datacubes

• msp_HSI_buteo.m

function cube = msp_HSI_buteo(file_path)
% msp_HSI_buteo - Reads and reshapes a hyperspectral image cube from a .pam file.
%
% This function reads a SWIR hyperspectral image stored in a .pam file and 
% reshapes it into a 3D cube format.
%
% Author: msp
% Date: February 2025
%
% Input:
%   file_path - String specifying the path to the .pam file.
%
% Output:
%   cube      - A 3D hyperspectral image cube (height x width x bands).
%
% Note:
%   - The function assumes a specific file format where the first 11 lines 
%     form a header that is ignored.
%   - The reshaping assumes a fixed image size of 1296 (height) × variable width × 900 (bands).
%

    % Open file for reading
    fileID = fopen(file_path, 'r');
    if fileID == -1
        error('Error: Could not open file %s', file_path);
    end

    % Read and store the first 11 lines as header
    header = [];
    for i = 1:11
        header = [header, fgets(fileID)]; % Read header, one line at a time
    end
    byte_length = numel(header); % Byte length of header (not used)

    % Read the rest of the .pam file (excluding header)
    data = uint8(fread(fileID)); % Read as unsigned 8-bit integer
    
    % Close the file
    fclose(fileID);
    
    % Reshape data into a 3D hyperspectral cube
    cube = reshape(data, 1296, [], 900); % SWIR Buteo assumed dimensions
    
    % Clear unnecessary variable
    clear data;
end

PAM interleave

The above example expects PAM files captured using the Buteo Hyperspectral Imaging System which have a BSQ interleave type.

Note that the PAM format native interleave is BIP.

Reflectance transformation

• msp_refl_trans.m

function cube_refl = msp_refl_trans(cube, mean_white, mean_dark)
% msp_refl_trans - Performs reflectance transformation on a hyperspectral image cube.
%
% This function converts raw hyperspectral intensity values into reflectance values
% using a white reference and a dark reference.
%
% Author: msp
% Date: February 2025
%
% Input:
%   cube       - A 3D hyperspectral image cube (height x width x bands).
%   mean_white - A reference white image (same dimensions as cube or a single spectrum).
%   mean_dark  - A reference dark image (same dimensions as cube or a single spectrum).
%
% Output:
%   cube_refl  - The hyperspectral cube transformed to reflectance values.

    % Compute reflectance transformation
    cube_refl = single(cube - uint8(mean_dark)) ./ single(uint8(mean_white - mean_dark));

    % Handle NaN and Inf values (caused by division by zero)
    cube_refl(isnan(cube_refl) | isinf(cube_refl)) = 0; 
    cube_refl = single(cube_refl); 
end