| Line 47... |
Line 47... |
| 47 |
% - random subsampling inspired by "Randomized Hough Transform for Ellipse Detection with Result Clustering"
|
47 |
% - random subsampling inspired by "Randomized Hough Transform for Ellipse Detection with Result Clustering"
|
| 48 |
% (CA Basca, M Talos, R Brad / 2005)
|
48 |
% (CA Basca, M Talos, R Brad / 2005)
|
| 49 |
%
|
49 |
%
|
| 50 |
% Update log:
|
50 |
% Update log:
|
| 51 |
% --------
|
51 |
% --------
|
| - |
|
52 |
% 1.2: Added more efficient distance calculation, less memory-usage, non-maximum suppression.
|
| 52 |
% 1.1: More memory efficient code, better documentation, more parameters, more solutions possible, example code.
|
53 |
% 1.1: More memory efficient code, better documentation, more parameters, more solutions possible, example code.
|
| 53 |
% 1.0: Initial version
|
54 |
% 1.0: Initial version
|
| 54 |
%
|
55 |
%
|
| 55 |
%
|
56 |
%
|
| 56 |
% Author: Martin Simonovsky
|
57 |
% Author: Martin Simonovsky
|
| Line 172... |
Line 173... |
| 172 |
weights = img(sub2ind(size(img),Y(K),X(K)));
|
173 |
weights = img(sub2ind(size(img),Y(K),X(K)));
|
| 173 |
end
|
174 |
end
|
| 174 |
accumulator = accumarray(idxs, weights, [params.maxMajorAxis 1]);
|
175 |
accumulator = accumarray(idxs, weights, [params.maxMajorAxis 1]);
|
| 175 |
|
176 |
|
| 176 |
%a bit of smoothing and finding the most busy bin
|
177 |
%a bit of smoothing and finding the most busy bin
|
| 177 |
%accumulator = conv(accumulator,H,'same');
|
178 |
accumulator = conv(accumulator,H,'same');
|
| 178 |
accumulator(1:ceil(sqrt(aSq)*params.minAspectRatio)) = 0;
|
179 |
accumulator(1:ceil(sqrt(aSq)*params.minAspectRatio)) = 0;
|
| 179 |
[score, idx] = max(accumulator);
|
180 |
[score, idx] = max(accumulator);
|
| 180 |
|
181 |
|
| 181 |
%keeping only the params.numBest best hypothesis (no non-maxima suppresion)
|
182 |
%keeping only the params.numBest best hypothesis
|
| 182 |
if (bestFits(end,end) < score)
|
183 |
if (bestFits(end,end) < score)
|
| - |
|
184 |
%non-maximum suppression (based on center distance for now)
|
| - |
|
185 |
squaredDists = (bestFits(:,1)-x0).^2 + (bestFits(:,2)-y0).^2;
|
| - |
|
186 |
[minSqDist, minSqDistIdx] = min(squaredDists);
|
| - |
|
187 |
if(minSqDist < 20^2 && score > bestFits(minSqDistIdx, 6))
|
| - |
|
188 |
bestFits(minSqDistIdx,:) = [x0 y0 sqrt(aSq) (idx-1) atand((y1-y2)/(x1-x2)) score];
|
| - |
|
189 |
elseif(minSqDist >= 10^2)
|
| 183 |
bestFits(end,:) = [x0 y0 sqrt(aSq) (idx-1) atand((y1-y2)/(x1-x2)) score];
|
190 |
bestFits(end,:) = [x0 y0 sqrt(aSq) (idx-1) atand((y1-y2)/(x1-x2)) score];
|
| - |
|
191 |
end
|
| 184 |
if params.numBest>1
|
192 |
if params.numBest>1
|
| 185 |
[~,si]=sort(bestFits(:,end),'descend');
|
193 |
[~,si]=sort(bestFits(:,end),'descend');
|
| 186 |
bestFits = bestFits(si,:);
|
194 |
bestFits = bestFits(si,:);
|
| 187 |
end
|
195 |
end
|
| 188 |
end
|
196 |
end
|