代碼
%%// Tolerance in percentage for the outliers/noise in the image because
%%// of which the edges are not perfectly vertical or horizontal and
%%// the ovalish blobs are not "round" enough
f=2;
%%// Read in your image
img = im2bw(imread('patt1.png'));
%%// Main processing
sum1 = sum(img,1);
box_startx = find(sum1>0.33*size(img,1),1);
box_stopx = size(img,2) - find(fliplr(sum1)>0.33*size(img,1),1) + 1;
sum2 = sum(img,2)'; %'
box_starty = find(sum2>0.33*size(img,2),1);
box_stopy = size(img,1) - find(fliplr(sum2)>0.33*size(img,2),1) + 1;
blob_leftx = find(sum1>(1-0.01*f)*max(sum1),1);
blob_rightx = size(img,2) - find(fliplr(sum1)>(1-0.01*f)*max(sum1),1) + 1;
blob_topy = find(sum2>(1-0.01*f)*max(sum2),1);
blob_bottomy = size(img,1) - find(fliplr(sum2)>(1-0.01*f)*max(sum2),1) + 1;
top1 = img(1:blob_topy,box_startx+1:box_stopx);
left1 = img(box_starty:box_stopy-1,1:blob_leftx);
bottom1 = img(blob_bottomy:end,box_startx:box_stopx);
right1 = img(box_starty:box_stopy,blob_rightx:end);
%// Debug
figure,
subplot(2,2,1);imshow(top1)
subplot(2,2,2);imshow(bottom1)
subplot(2,2,3);imshow(left1)
subplot(2,2,4);imshow(right1)
輸出
這是你的真實形象?那麼你的形象真的是黑色和白色?因爲如果是這樣,就像找到直線的交點一樣簡單,然後找到相應子陣列中的最後/第一個白色像素;無需訴諸功能檢測。 –
哦,哇...我會試試,但Divakar的方法功能完美。謝謝你的提示;深夜編碼會讓人頭腦發熱! – user1106340
@ user1106340我想這就是我所做的,簡單的數學,沒有複雜的措施。 – Divakar