Image Segmentation using their magnitude and direction gradients

Hello all, i have attacthed magnitude and direction gradients of 3 images. I am trying to achieve a segmentation where the section below the red curve is removed and the section above the red curve should be preserved. I have shown below for image 2:

I have attatched what i want to acheive for image 1 and image 3 as Image-1 and image-3. I used paint to highlight the red curve. Can you guys suggest me some ways to acheive this? Any help is appreciated and thankyou in advance. The code i used is below

clc;
clear;
close all;
a=imread(filename);
B = imgaussfilt(a,1.4);
[Gmag,Gdir]=imgradient(B,'prewitt');
figure;
imshowpair(Gmag,Gdir,'montage');

I have also attacthed the original images as org1, org2 and org3.

Answer 1

[Image Analyst]

Just threshold the image and use find() to find the last row in each column where the signal is above the threshold.

thresholdValue = 10; % Whatever works
[rows, columns, numberOfColorChannels] = size(grayImage);
lastRows = zeros(1, columns);
for col = 1 : columns
    r = find(grayImage(:, col) > threshold, 1, 'last')
    if ~isempty(r)
        lastRows(col) = r;
    end
end

[Apoorva Maiya]

Thank you for your help but this methos doesnt seem to work since some rows are empty. Is there anything we could do with the direction gradient because it shows a sharp ridge at whare I want to segment the image.