请求讲解这段grabcut算法的核心代码
function [] = SegmentGC(ImName, ih, alg)
% SEGMENTGC - Main GrabCut segmentation function
% SEGMENTGC(IMNAME, IH, ALG) - ImName - Image to segment
% IH - Image Handle
% ALG - AutoCut or AutoRefine (0 or 1)
% Authors - Mohit Gupta, Krishnan Ramnath
% Affiliation - Robotics Institute, CMU, Pittsburgh
% 2006-05-15
% Global Variables
global sopt segImageHandle;
% Read input image
I = imread(ImName);
I1 = I(:,:,1);
I2 = I(:,:,2);
I3 = I(:,:,3);
% Final Segmented Image
SegImage = zeros(size(I));
% Stop unnecessary warnings
warning('off','all');
%%%%%%% Marking the Inner Rectangle around the foreground object
%%%%%%% Get two points from the user
% Set the image
set(ih, 'Cdata', I);
axis('image');axis('ij');axis('off');
drawnow;
fp = ginput(2);
fp = round(fp);
% Save the points for future use
save rectpts fp;
% Width of Strip
StripWidth = sopt.StripWidth_AC_ACR;
% Get inner indices
[fxi,fyi] = meshgrid(min(fp(:,1)):max(fp(:,1)),min(fp(:,2)):max(fp(:,2)));
InsideFindices = sub2ind(size(I1),fyi, fxi);
InsideFindices = InsideFindices(:);
% Form the Outside rectangular bounding box -- add StripWidth on all 4 sides
[fxo,fyo] = meshgrid( max((min(fp(:,1)) - StripWidth),1) : min((max(fp(:,1)) + StripWidth), size(I1,2)) , max((min(fp(:,2)) - StripWidth),1) : min((max(fp(:,2)) + StripWidth), size(I1,1)) );
OutsideFindices = sub2ind(size(I1),fyo, fxo);
OutsideFindices = OutsideFindices(:);
Bindices = setdiff(OutsideFindices, InsideFindices);
Allindices = [1:size(I1,1)*size(I1,2)]';
Findices = InsideFindices;
% Water shed pixels
L = watershed(I(:,:,1));
%%%%%%%%% Finding Mean colors of the regions %%%%%%%%%%%%%%%%%%
numLabels = max(L(:));
PI = regionprops(L,'PixelIdxList');
MeanColors = zeros(numLabels,3);
for i=1:numLabels
MeanColors(i,1) = mean(I1(PI(i).PixelIdxList));
MeanColors(i,2) = mean(I2(PI(i).PixelIdxList));
MeanColors(i,3) = mean(I3(PI(i).PixelIdxList));
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%% Finding Foreground and Background Labels from pixels%%
FLabels = L(int32(Findices));
FLabels = union(FLabels,[]); %% Set-ifying the set of labels (Sorting)
if(FLabels(1)==0) %% Removing Boundary Labels
FLabels = FLabels(2:end);
end
BLabels = L(int32(Bindices));
BLabels = union(BLabels,[]); %% Set-ifying the set of labels
if(BLabels(1)==0) %% Removing Boundary Labels
BLabels = BLabels(2:end);
end
%%%% BLabels remain fixed -- no change
%%%% ULabels -- Uncertain labels -- on which optimization done...
ULabels = FLabels;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% Finding Initial Estimate of foregound and background color clusters %%%%%%%%%%%%
NumFClusters = sopt.NumFClusters_AC;
NumBClusters = sopt.NumBClusters_AC;
FColors = MeanColors(FLabels,:);
BColors = MeanColors(BLabels,:);
%%%%%%%%%%% Initializing the GMM's %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%% Using Just kmeans %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[FId FCClusters] = kmeans(FColors, NumFClusters);
Fdim = size(FColors,2);
FCClusters = zeros(Fdim, NumFClusters);
FWeights = zeros(1,NumFClusters);
FCovs = zeros(Fdim, Fdim, NumFClusters);
for k=1:NumFClusters
relColors = FColors(find(FId==k),:); %% Colors belonging to cluster k
FCClusters(:,k) = mean(relColors,1)';
FCovs(:,:,k) = cov(relColors);
FWeights(1,k) = length(find(FId==k)) / length(FId);
end
[BId BCClusters] = kmeans(BColors, NumBClusters);
Bdim = size(BColors,2);
BCClusters = zeros(Bdim, NumBClusters);
BWeights = zeros(1,NumBClusters);
BCovs = zeros(Bdim, Bdim, NumBClusters);
for k=1:NumBClusters
relColors = BColors(find(BId==k),:); %% Colors belonging to cluster k
BCClusters(:,k) = mean(relColors,1)';
BCovs(:,:,k) = cov(relColors);
BWeights(1,k) = length(find(BId==k)) / length(BId);
end
%%%%%%%%%%%% GMM's Initialized %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%% The iterative Loop %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
lambda = sopt.lambda_AC;
numIter = sopt.numIter_AC;
for i = 1:numIter
disp(['Iteration - ', num2str(i)]);
% Foreground and Background edge weights
[FDist, FInd] = ClustDistMembership(MeanColors(ULabels,:), FCClusters, FCovs, FWeights);
[BDist, BInd] = ClustDistMembership(MeanColors(ULabels,:), BCClusters, BCovs, BWeights);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Segments labeled from 0 now -- -1 is for boundary pixels
L = L-1;
FLabels = FLabels-1;
BLabels = BLabels-1;
ULabels = ULabels-1;
%%%%%%%%%%%%% The Mex Function for GraphCutSegment %%%%%%%%%%%%
%%% SegImage is the segmented image, %%% LLabels is the binary label
%%% for each watershed label
[SegImage LLabels] = GraphCutSegment(L, MeanColors, ULabels, BLabels,...
FDist, BDist, lambda);
%% Again Labeled from 1...
L = L+1;
FLabels = FLabels+1;
BLabels = BLabels+1;
ULabels = ULabels+1;
if(i < numIter)
%%%%%% Do NOT do this if final iteration -- just display the segmented image
%%%%%% Making new FLabels and BLabels based on the segmentation %%%%%%
newFLabels = ULabels(find(LLabels==1.0));
newBLabels = ULabels(find(LLabels==0.0));
%%%%%% Whether new background labels will contain the old ones?
% newBLabels = union(newBLabels,BLabels);
FColors = MeanColors(newFLabels,:);
BColors = MeanColors(newBLabels,:);
%%%%%%%% Calculating FG and BG distances based on new segmentation %%%%%%%%%%%%%%%%%%
[newFDists newFInd] = ClustDistMembership(FColors, FCClusters, FCovs, FWeights);
[newBDists newBInd] = ClustDistMembership(BColors, BCClusters, BCovs, BWeights);
for k = 1:NumFClusters
relColors = FColors(find(newFInd==k),:); %% Colors belonging to cluster k
FCClusters(:,k) = mean(relColors,1)';
FCovs(:,:,k) = cov(relColors);
FWeights(1,k) = length(find(newFInd==k)) / length(newFInd);
end
for k=1:NumBClusters
relColors = BColors(find(newBInd==k),:); %% Colors belonging to cluster k
BCClusters(:,k) = mean(relColors,1)';
BCovs(:,:,k) = cov(relColors);
BWeights(1,k) = length(find(newBInd==k)) / length(newBInd);
end
end
end
%%%%%%%%%%%%%%%%%%%%% Display the segmented image %%%%%%%%%%%%%%%%%%%
edge_img = edge(SegImage,'canny');
% Put image on black background
SegImage = repmat(SegImage,[1,1,3]);
SegNewImage = uint8(SegImage) .* uint8(I);
% If AutoCut, just show the image
if(~alg)
figure;
imshow(uint8(SegNewImage));
else
% If AutoRefine mark a segmentation boundary on original image
[IInd,JInd] = ind2sub(size(I1),find(edge_img));
boundImage1 = I(:,:,2);
boundImage1(find(edge_img)) = 255;
boundImage = I;
boundImage(:,:,2) = boundImage1;
% Set the image
set(ih, 'Cdata', uint8(boundImage));
axis('image');axis('ij');axis('off');
drawnow;
segImageHandle = figure;
imshow(uint8(SegNewImage));
end
SegMask = SegImage;
SegResult = SegNewImage;
% Save Segmentation Result
save('SegResult', 'SegMask', 'SegResult');
% Required for AutoCutRefine
save('iter_data','L', 'MeanColors', 'FLabels', 'BLabels', 'FCClusters',...
'FCovs', 'FWeights', 'BCClusters', 'BCovs', 'BWeights');
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%% Helper Functions declarations %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [FDist, FInd] = ClustDistMembership(MeanColors, FCClusters, FCovs, FWeights)
% CLUSTDISTMEMBERSHIP - Calcuates FG and BG Distances
% Authors - Mohit Gupta, Krishnan Ramnath
% Affiliation - Robotics Institute, CMU, Pittsburgh
% 2006-05-15
NumFClusters = size(FCClusters,2);
numULabels = size(MeanColors,1);
FDist = zeros(numULabels,1);
FInd = zeros(numULabels,1);
Ftmp = zeros(numULabels, NumFClusters);
for k = 1:NumFClusters
M = FCClusters(:,k);
CovM = FCovs(:,:,k);
W = FWeights(1,k);
V = MeanColors - repmat(M',numULabels,1);
Ftmp(:,k) = -log((W / sqrt(det(CovM))) * exp(-( sum( ((V * inv(CovM)) .* V),2) /2)));
end
[FDist, FInd] = min(Ftmp,[],2);
您好，我是有问必答小助手，您的提问已进入加速通道，请您耐心等待技术专家团成员的回复~
C和C++完整教程：https://blog.csdn.net/it_xiangqiang/category_10581430.html
您好，我是有问必答小助手，你的问题已经有小伙伴为您解答了问题，您看下是否解决了您的问题，可以追评进行沟通哦~
如果有您比较满意的答案 / 帮您提供解决思路的答案，可以点击【采纳】按钮，给回答的小伙伴一些鼓励哦～～
ps:问答VIP仅需29元，即可享受5次/月 有问必答服务，了解详情>>>https://vip.csdn.net/askvip?utm_source=1146287632