本帖最后由 fuyunchao 于 2018-7-25 16:53 编辑

在运行一个FABEMD的程序，出现以下报错，其中se_rep是fabemd的一个子函数。网上也有很多类似的问题，但是程序不一样也看不懂，希望专家或者朋友或者老师们帮忙看看，谢谢

已达到最大递归限制 500。使用 set(0,'RecursionLimit',N) 可更改此限制。请注意，超出可用堆栈空间可能会使 MATLAB 和/或计算机崩溃。

出错 se_rep

运行程序如下：

A=imread('childern.tif');

H=fabemd(A);

出错函数是：

function r = se_rep(SE)

% se_rep - representation type of structuring element

% ---------------------------------------------------

%

% r = se_rep(SE)

%

% Input:

% ------

%  SE - structuring element

%

% Output:

% -------

%  r - representation type 'mat', 'vec', or ''

%--------------------------------

% Author: Harold Figueroa

%--------------------------------

% $Date: 2006-06-06 17:29:13 -0400 (Tue, 06 Jun 2006) $

% $Revision: 5169 $

%--------------------------------

[m, n] = size(SE);

if (n == 2) && all(round(SE(:)) == SE(:))

r = 'vec';

elseif rem(m, 2) && rem(n, 2) && all((SE(:) ~= 0) == SE(:))

r = 'mat';

else

r = ''; % not a structuring element

if ~nargout

warning('Input is not a structuring element.');

end

end

fabemd函数为：

function Y = fabemd(X, opt)

% fabemd - fast adaptive bidimensional EMD

% ----------------------------------------

%

% Y = fabemd(X, opt)

%

% Input:

% ------

%  X - image

%  opt - options

%

% Output:

% -------

%  Y - BIMF array

% TODO: save intermediate computations, specifically extrema

% NOTE: consider a single coded image for extrema across iterations

%--

% handle input

%--

if nargin < 2

opt.type = 4;

opt.radial = false;

opt.verb = 1;

if ~nargin && nargout

Y = opt; return;

end

end

if ~nargout || opt.verb

figure; imagesc(X), colorbar; colormap(gray(256)); axis('image');

end

for k = 1:3

[L, U] = get_envelopes(X, opt);

Y(:, :, k) = X - 0.5 * (L + U);

X = X - Y(:, :, k);

end

Y(:, :, end + 1) = X;

if ~nargout || opt.verb

for k = 1:size(Y, 3), figure; imagesc(Y(:,:,k)), colorbar; cmap_real(129); axis image; end

end

%-----------------------

% GET_ENVELOPES

%-----------------------

% NOTE: this is essentially a method of the EMD computation

function [L, U, E] = get_envelopes(X, opt)

% get_envelopes - for image

% -------------------------

%

% [L, U, E] = get_envelopes(X, opt)

%

% Input:

% ------

%  X - image

%  opt - options, coming from parent

%

% Output:

% -------

%  L, U - lower and upper envelopes

%  E - extrema indicator

%--

% compute image extrema

%--

[E, count] = image_extrema(X); %#ok

% TODO: consider count before we move on

%--

% compute extrema distances and radius of operations given type

%--

[p(:, 1), p(:, 2)] = find(E == 1, 2000);

pd = fast_nearest(p'); pd = pd(:, 1);

[v(:, 1), v(:, 2)] = find(E == -1, 2000);

vd = fast_nearest(v'); vd = vd(:, 1);

switch opt.type

case 1 % min-min

r = min(min(pd), min(vd));

case 2 % max-min

r = max(min(pd), min(vd));

case 3 % min-max

r = min(max(pd), max(vd));

case 4 % max-max

r = max(max(pd), max(vd));

end

r = next_odd(r);

%--

% compute envelopes

%--

% NOTE: we should be able to make use of CUDA acceleration for all steps below

% TODO: improve filtering, improve filter

if opt.radial

SE = se_ball(r); F = SE / se_size(SE);

else

SE = ones(r); F = SE / r^2;

end

% NOTE: 'morph' functions decompose the structuring element, below we decompose the linear filter

L = morph_erode(X, SE); U = morph_dilate(X, SE);

L = box_filter(L, F) / r^2;

U = box_filter(U, F) / r^2;