decision_function.m

function nsim_score = compute_nsim(neurogram_ref, neurogram_degraded, C1, C2)
    % Compute the neurogram similarity index measure (NSIM).
    if nargin < 3
        C1 = 0.01;
    end
    if nargin < 4
        C2 = 0.03;
    end

    neurogram_ref = normalize(neurogram_ref);
    neurogram_degraded = normalize(neurogram_degraded);

    mu_r = mean(neurogram_ref(:));
    mu_d = mean(neurogram_degraded(:));
    sigma_r = var(neurogram_ref(:)) + 1e-10;
    sigma_d = var(neurogram_degraded(:)) + 1e-10;
    sigma_rd = cov(neurogram_ref(:), neurogram_degraded(:));
    sigma_rd = sigma_rd(1, 2);

    % Check if variances are zero and handle the edge case
    if sigma_r <= 1e-10 && sigma_d <= 1e-10
        nsim_score = 0;
        return;
    end

    % Compute luminance term
    luminance = (2 * mu_r * mu_d + C1) / (mu_r^2 + mu_d^2 + C1);

    % Compute contrast term
    contrast = (sigma_rd + C2) / (sqrt(sigma_r * sigma_d) + C2);

    nsim_score = luminance * contrast;

    % Clamp the NSIM score to the range [0, 1] to avoid slight deviations above 1
    nsim_score = min(max(nsim_score, 0), 1);
end

function normalized_neurogram = normalize(neurogram)
    % Normalize the neurogram to the range [0, 1].
    min_val = min(neurogram(:));
    max_val = max(neurogram(:));
    if max_val - min_val == 0
        normalized_neurogram = zeros(size(neurogram));
    else
        normalized_neurogram = (neurogram - min_val) / (max_val - min_val);
    end
end

% Example usage
neurogram_ref = rand(100, 50); % example reference neurogram
neurogram_degraded = rand(100, 50); % example degraded neurogram

% compute NSIM score
nsim_score = compute_nsim(neurogram_ref, neurogram_degraded);
fprintf('NSIM Score: %.4f\n', nsim_score);

% Example usage
neurogram_ref = ones(100, 50);
neurogram_degraded = zeros(100, 50);

nsim_score = compute_nsim(neurogram_ref, neurogram_degraded);
fprintf('NSIM Score: %.4f\n', nsim_score);