look for submarines using Hough transform

src/MATLAB/+Segmentation/FrameSegmentor.m

@@ -258,4 +258,66 @@ function [objs features levels] = FrameSegmentor(im, t, imageAlpha)
         objs=[objs no];
         features = [features nf];
     end
+    
+    % hemato submarines
+%    [sharpImg blurImg bw] = michelContrastEnhance(1-im);
+    [bw] = Segmentation.Michel(1-im, [3 3]);
+    
+    [centers, radii, metric] = imfindcircles(bw, [10 20], 'Sensitivity', 0.9);
+    [r c] = find(bw >= 0);
+    for i=1:length(radii)
+        % find the pixels inside the circle
+        dist = ((c - centers(i,1)).^2 + (r - centers(i,2)).^2);
+        pix = find(dist <= radii(i)^2);
+        [r1 c1] = ind2sub(size(im), pix);
+        
+        % find the convex hull of the circle
+        ch = convhull(c1, r1);
+
+        no = [];
+        no.t = t;
+        no.points = [c(ch), r(ch)];
+        no.ID = -1;
+        no.indPixels = pix;
+        no.indTailPixels  = [];
+        no.BrightInterior=1;
+        no.Eccentricity=0;
+        no.imPixels=im(pix);
+        
+        nf = [];
+        nf.darkRatio = 0;
+        nf.haloRatio = 0;
+        nf.igRatio = 0;
+        nf.darkIntRatio = 0;
+        nf.brightInterior = 1;
+        
+        nf.polyPix = [];
+        nf.perimPix = [];
+        nf.igPix = [];
+        nf.haloPix = [];
+        
+        objs=[objs no];
+        features = [features nf];
+    end
+end
+
+function [sharpImg blurImg bw] = michelContrastEnhance(img)
+
+    bluriter = 20;
+    blurradius = 15;
+    blurstd = 5;
+
+    smoothFilt = fspecial('gaussian', 2*blurradius*(bluriter)+1, sqrt(bluriter)*blurstd);
+
+    blurImg = imfilter(img, smoothFilt, 'symmetric');
+    sharpImg = max(0, img - blurImg);
+
+    % median filter the high-frequency image
+    medim = medfilt2(sharpImg,[5 5]);
+
+    % The rest of this code removes purely zero values then uses an otsu threshold to segment the image
+    medimZ = medim(medim>0);
+
+    th = graythresh(medimZ);
+    bw = im2bw(medim,th);
 end