matlab call Ncut: Matrix is too large to convert to linear index.

在调用Ncut的matlab代码的时候发现出现如下问题：

Matrix is too large to convert to linear index.

原始的错误信息如下：

??? Matrix is too large to convert to linear index.

Error in ==> computeW at 10
W = W/max(W(:));

Error in ==> ICgraph at 37
W = computeW(I,dataW,edgemap.emag,edgemap.ephase);

Error in ==> NcutImage at 16
[W,imageEdges] = ICgraph(I);

Error in ==> get_segment at 29
    [SegLabel,NcutDiscrete,NcutEigenvectors,NcutEigenvalues,W,imageEdges]=
    NcutImage(I,nbSegments);

Error in ==> demo_aborn at 11
map = get_segment(imageName,nbSegments,1);

定位到出错的函数：

function W = computeW(imageX,dataW,emag,ephase)
% W = computeW(imageX,dataW,emag,ephase)
% Timothee Cour, Stella Yu, Jianbo Shi, 2004.
[p,q] = size(imageX);

[w_i,w_j] = cimgnbmap([p,q],dataW.sampleRadius,dataW.sample_rate);

W = affinityic(emag,ephase,w_i,w_j,max(emag(:)) * dataW.edgeVariance);
W = W/max(W(:));

错误出现在最后一行的max上，我发现如果用尺寸较小的image做实验不会出现这个问题：

估计是求最大值max(W(:))在矩阵太大的时候会出bug的原因

把原代码改成如下就行了：

function W = computeW(imageX,dataW,emag,ephase)
% W = computeW(imageX,dataW,emag,ephase)
% Timothee Cour, Stella Yu, Jianbo Shi, 2004.
[p,q] = size(imageX);

[w_i,w_j] = cimgnbmap([p,q],dataW.sampleRadius,dataW.sample_rate);

W = affinityic(emag,ephase,w_i,w_j,max(emag(:)) * dataW.edgeVariance);
W = W/max(max(W));