Very fast 2D local maxima

Detects local maxima (peaks) in a 2D image above a specified intensity threshold, neighborhood, and optionally filters peaks by prominence

A very fast local maxima (peak detector) for 2D images using morphological maximum filtering. Ideal for video analysis when used in a loop.

User can define intensity threshold, local neighborhood and optionally filters peaks by prominence.

USAGE:

Example 1 (defult): locs = Fast_peaks2D(Img, thresh, kernel_size)

Example 2 (with prominence): locs = Fast_peaks2D(Img, thresh, kernel_size, min_prom)

Example 3 (for multiple frames):

locs = [];

for i = 1:numel(img(1,1,:))

li = Fast_peaks2D(img(:,:,i), LAFM_thresh, 1);

li = [li, i*ones(numel(li(:,1)),1)];

locs = [locs; li];

end

INPUTS:

Img - 2D image (grayscale, numeric matrix) for peak detection.

thresh - Intensity threshold; only peaks above this value are considered.

kernel_size - Size of the local neighborhood used for maximum filtering (in pixels).

min_prom - (Optional) Minimum required prominence of peaks. Peaks with lower prominence are discarded.

OUTPUT:

locs - Nx4 array where each row represents a detected peak:

[x_position, y_position, peak_height, prominence]

If min_prom is not provided, prominence is set to zero.

ISSUE: Over picks peaks on featureless or uniform regions

NOTES:

• The function uses morphological maximum filtering and optional prominence filtering.
• Edge pixels (within 2 pixels of the image boundary) are excluded from detection.
• Prominence is computed as the vertical distance to the lowest point along the path to the nearest higher neighboring peak. Prominence calculations slow the code down.