如何在 Python 中像 edge() Matlab 函数一样制作精巧的边缘检测器
Posted
技术标签:
【中文标题】如何在 Python 中像 edge() Matlab 函数一样制作精巧的边缘检测器【英文标题】：How to make a canny edge detector in Python like edge() Matlab function 【发布时间】：2018-08-12 13:48:45 【问题描述】：我想在 Python 中使用 edge() Matlab 函数：
mask = double(edge(mask,'canny',threshold));
我正在尝试在 Python 中找到一个与 edge() Matlab 函数等效的函数。我在 Python 中尝试了 cv2.Canny() 函数，但它与 Matlab 函数不同（您可以在此处阅读关于差异的解释：https://dsp.stackexchange.com/questions/4716/differences-between-opencv-canny-and-matlab-canny）。我已经尝试过该论坛中的建议，首先使用高斯模糊来模糊图像，但结果仍然与 matlab 不同。
所以，我找到了一些 Matlab 精明的边缘检测器代码，我正试图将这些代码转换为内置的 edge() 函数，就像 Matlab 的一样。我找到的代码是：
第一个代码：
clear all;
clc;
%Input image
img = imread ('gambar.jpg');
%Show input image
figure, imshow(img);
img = rgb2gray(img);
img = double (img);
%Value for Thresholding
T_Low = 0.2;
T_High = 0.5;
%Gaussian Filter Coefficient
B = [2, 4, 5, 4, 2; 4, 9, 12, 9, 4;5, 12, 15, 12, 5;4, 9, 12, 9, 4;2, 4, 5, 4, 2 ];
B = 1/159.* B;
%Convolution of image by Gaussian Coefficient
A=conv2(img, B, 'same');
%Filter for horizontal and vertical direction
KGx = [-1, 0, 1; -2, 0, 2; -1, 0, 1];
KGy = [1, 2, 1; 0, 0, 0; -1, -2, -1];
%Convolution by image by horizontal and vertical filter
Filtered_X = conv2(A, KGx, 'same');
Filtered_Y = conv2(A, KGy, 'same');
%Calculate directions/orientations
arah = atan2 (Filtered_Y, Filtered_X);
arah = arah*180/pi;
pan=size(A,1);
leb=size(A,2);
%Adjustment for negative directions, making all directions positive
for i=1:pan
for j=1:leb
if (arah(i,j)<0)
arah(i,j)=360+arah(i,j);
end;
end;
end;
arah2=zeros(pan, leb);
%Adjusting directions to nearest 0, 45, 90, or 135 degree
for i = 1 : pan
for j = 1 : leb
if ((arah(i, j) >= 0 ) && (arah(i, j) < 22.5) || (arah(i, j) >= 157.5) && (arah(i, j) < 202.5) || (arah(i, j) >= 337.5) && (arah(i, j) <= 360))
arah2(i, j) = 0;
elseif ((arah(i, j) >= 22.5) && (arah(i, j) < 67.5) || (arah(i, j) >= 202.5) && (arah(i, j) < 247.5))
arah2(i, j) = 45;
elseif ((arah(i, j) >= 67.5 && arah(i, j) < 112.5) || (arah(i, j) >= 247.5 && arah(i, j) < 292.5))
arah2(i, j) = 90;
elseif ((arah(i, j) >= 112.5 && arah(i, j) < 157.5) || (arah(i, j) >= 292.5 && arah(i, j) < 337.5))
arah2(i, j) = 135;
end;
end;
end;
figure, imagesc(arah2); colorbar;
%Calculate magnitude
magnitude = (Filtered_X.^2) + (Filtered_Y.^2);
magnitude2 = sqrt(magnitude);
BW = zeros (pan, leb);
%Non-Maximum Supression
for i=2:pan-1
for j=2:leb-1
if (arah2(i,j)==0)
BW(i,j) = (magnitude2(i,j) == max([magnitude2(i,j), magnitude2(i,j+1), magnitude2(i,j-1)]));
elseif (arah2(i,j)==45)
BW(i,j) = (magnitude2(i,j) == max([magnitude2(i,j), magnitude2(i+1,j-1), magnitude2(i-1,j+1)]));
elseif (arah2(i,j)==90)
BW(i,j) = (magnitude2(i,j) == max([magnitude2(i,j), magnitude2(i+1,j), magnitude2(i-1,j)]));
elseif (arah2(i,j)==135)
BW(i,j) = (magnitude2(i,j) == max([magnitude2(i,j), magnitude2(i+1,j+1), magnitude2(i-1,j-1)]));
end;
end;
end;
BW = BW.*magnitude2;
figure, imshow(BW);
%Hysteresis Thresholding
T_Low = T_Low * max(max(BW));
T_High = T_High * max(max(BW));
T_res = zeros (pan, leb);
for i = 1 : pan
for j = 1 : leb
if (BW(i, j) < T_Low)
T_res(i, j) = 0;
elseif (BW(i, j) > T_High)
T_res(i, j) = 1;
%Using 8-connected components
elseif ( BW(i+1,j)>T_High || BW(i-1,j)>T_High || BW(i,j+1)>T_High || BW(i,j-1)>T_High || BW(i-1, j-1)>T_High || BW(i-1, j+1)>T_High || BW(i+1, j+1)>T_High || BW(i+1, j-1)>T_High)
T_res(i,j) = 1;
end;
end;
end;
edge_final = uint8(T_res.*255);
%Show final edge detection result
figure, imshow(edge_final);
来自： https://www.mathworks.com/matlabcentral/mlc-downloads/downloads/submissions/63294/versions/9/previews/MaxPol%20Package/demo%20examples/forward%20signal%20and%20imaging%20problems/Canny%20edge%20detection/canny_edge.m/index.html?access_key=
第二个代码：
function [eout, dx, dy] = canny_edge(image_scan, smoothing_kernel, derivative_kernel)
thresh = [0.2 0.5];
sigma = sqrt(2);
thinning = true;
H = [];
kx = 1;
ky = 1;
[m,n] = size(image_scan);
e = false(m,n);
% Magic numbers
PercentOfPixelsNotEdges = .7; % Used for selecting thresholds
ThresholdRatio = .4; % Low thresh is this fraction of the high.
dx = imfilter(image_scan, smoothing_kernel', 'conv', 'replicate');
dx = imfilter(dx, derivative_kernel, 'conv', 'replicate');
% Compute smoothed numerical gradient of image I along y (vertical)
% direction. GY corresponds to dG/dy, where G is the Gaussian Smoothed
% version of image I.
dy = imfilter(image_scan, smoothing_kernel, 'conv', 'replicate');
dy = imfilter(dy, derivative_kernel', 'conv', 'replicate');
% Calculate Magnitude of Gradient
magGrad = hypot(dx, dy);
% Normalize for threshold selection
magmax = max(magGrad(:));
if magmax > 0
magGrad = magGrad / magmax;
end
% Determine Hysteresis Thresholds
[lowThresh, highThresh] = selectThresholds(thresh, magGrad, PercentOfPixelsNotEdges, ThresholdRatio, mfilename);
% Perform Non-Maximum Suppression Thining and Hysteresis Thresholding of Edge
% Strength
eout = thinAndThreshold(e, dx, dy, magGrad, lowThresh, highThresh);
thresh = [lowThresh highThresh];
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : selectThresholds
%
function [lowThresh, highThresh] = selectThresholds(thresh, magGrad, PercentOfPixelsNotEdges, ThresholdRatio, ~)
[m,n] = size(magGrad);
% Select the thresholds
if isempty(thresh)
counts=imhist(magGrad, 64);
highThresh = find(cumsum(counts) > PercentOfPixelsNotEdges*m*n,...
1,'first') / 64;
lowThresh = ThresholdRatio*highThresh;
elseif length(thresh)==1
highThresh = thresh;
if thresh>=1
error(message('images:edge:thresholdMustBeLessThanOne'))
end
lowThresh = ThresholdRatio*thresh;
elseif length(thresh)==2
lowThresh = thresh(1);
highThresh = thresh(2);
if (lowThresh >= highThresh) || (highThresh >= 1)
error(message('images:edge:thresholdOutOfRange'))
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : thinAndThreshold
%
function H = thinAndThreshold(E, dx, dy, magGrad, lowThresh, highThresh)
% Perform Non-Maximum Suppression Thining and Hysteresis Thresholding of Edge
% Strength
% We will accrue indices which specify ON pixels in strong edgemap
% The array e will become the weak edge map.
idxStrong = [];
for dir = 1:4
idxLocalMax = cannyFindLocalMaxima(dir,dx,dy,magGrad);
idxWeak = idxLocalMax(magGrad(idxLocalMax) > lowThresh);
E(idxWeak)=1;
idxStrong = [idxStrong; idxWeak(magGrad(idxWeak) > highThresh)]; %#ok<AGROW>
end
[m,n] = size(E);
if ~isempty(idxStrong) % result is all zeros if idxStrong is empty
rstrong = rem(idxStrong-1, m)+1;
cstrong = floor((idxStrong-1)/m)+1;
H = bwselect(E, cstrong, rstrong, 8);
else
H = zeros(m, n);
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : cannyFindLocalMaxima
%
function idxLocalMax = cannyFindLocalMaxima(direction,ix,iy,mag)
%
% This sub-function helps with the non-maximum suppression in the Canny
% edge detector. The input parameters are:
%
% direction - the index of which direction the gradient is pointing,
% read from the diagram below. direction is 1, 2, 3, or 4.
% ix - input image filtered by derivative of gaussian along x
% iy - input image filtered by derivative of gaussian along y
% mag - the gradient magnitude image
%
% there are 4 cases:
%
% The X marks the pixel in question, and each
% 3 2 of the quadrants for the gradient vector
% O----0----0 fall into two cases, divided by the 45
% 4 | | 1 degree line. In one case the gradient
% | | vector is more horizontal, and in the other
% O X O it is more vertical. There are eight
% | | divisions, but for the non-maximum suppression
% (1)| |(4) we are only worried about 4 of them since we
% O----O----O use symmetric points about the center pixel.
% (2) (3)
[m,n] = size(mag);
% Find the indices of all points whose gradient (specified by the
% vector (ix,iy)) is going in the direction we're looking at.
switch direction
case 1
idx = find((iy<=0 & ix>-iy) | (iy>=0 & ix<-iy));
case 2
idx = find((ix>0 & -iy>=ix) | (ix<0 & -iy<=ix));
case 3
idx = find((ix<=0 & ix>iy) | (ix>=0 & ix<iy));
case 4
idx = find((iy<0 & ix<=iy) | (iy>0 & ix>=iy));
end
% Exclude the exterior pixels
if ~isempty(idx)
v = mod(idx,m);
extIdx = (v==1 | v==0 | idx<=m | (idx>(n-1)*m));
idx(extIdx) = [];
end
ixv = ix(idx);
iyv = iy(idx);
gradmag = mag(idx);
% Do the linear interpolations for the interior pixels
switch direction
case 1
d = abs(iyv./ixv);
gradmag1 = mag(idx+m).*(1-d) + mag(idx+m-1).*d;
gradmag2 = mag(idx-m).*(1-d) + mag(idx-m+1).*d;
case 2
d = abs(ixv./iyv);
gradmag1 = mag(idx-1).*(1-d) + mag(idx+m-1).*d;
gradmag2 = mag(idx+1).*(1-d) + mag(idx-m+1).*d;
case 3
d = abs(ixv./iyv);
gradmag1 = mag(idx-1).*(1-d) + mag(idx-m-1).*d;
gradmag2 = mag(idx+1).*(1-d) + mag(idx+m+1).*d;
case 4
d = abs(iyv./ixv);
gradmag1 = mag(idx-m).*(1-d) + mag(idx-m-1).*d;
gradmag2 = mag(idx+m).*(1-d) + mag(idx+m+1).*d;
end
idxLocalMax = idx(gradmag>=gradmag1 & gradmag>=gradmag2);
来自： https://www.mathworks.com/matlabcentral/fileexchange/46859-canny-edge-detection
我试图转换成这个内置的edge() matlab 函数：
function [eout,thresh,gv_45,gh_135] = edge(varargin)
args = matlab.images.internal.stringToChar(varargin);
[a,method,thresh,sigma,thinning,H,kx,ky] = parse_inputs(args:);
% Check that the user specified a valid number of output arguments
if ~any(strcmp(method,'sobel','roberts','prewitt')) && (nargout>2)
error(message('images:edge:tooManyOutputs'))
end
% Transform to a double precision intensity image if necessary
isPrewittOrSobel = strcmp(method,'sobel') || strcmp(method,'prewitt');
if ~isPrewittOrSobel && ~isfloat(a) && ~strcmp(method,'approxcanny')
a = im2single(a);
end
[m,n] = size(a);
if strcmp(method,'canny')
% Magic numbers
PercentOfPixelsNotEdges = .7; % Used for selecting thresholds
ThresholdRatio = .4; % Low thresh is this fraction of the high.
% Calculate gradients using a derivative of Gaussian filter
[dx, dy] = smoothGradient(a, sigma);
% Calculate Magnitude of Gradient
magGrad = hypot(dx, dy);
% Normalize for threshold selection
magmax = max(magGrad(:));
if magmax > 0
magGrad = magGrad / magmax;
end
% Determine Hysteresis Thresholds
[lowThresh, highThresh] = selectThresholds(thresh, magGrad, PercentOfPixelsNotEdges, ThresholdRatio, mfilename);
% Perform Non-Maximum Suppression Thining and Hysteresis Thresholding of Edge
% Strength
e = thinAndThreshold(dx, dy, magGrad, lowThresh, highThresh);
thresh = [lowThresh highThresh];
elseif strcmp(method,'approxcanny')
e = computeapproxcanny(a, thresh);
elseif strcmp(method,'canny_old')
% Magic numbers
GaussianDieOff = .0001;
PercentOfPixelsNotEdges = .7; % Used for selecting thresholds
ThresholdRatio = .4; % Low thresh is this fraction of the high.
% Design the filters - a gaussian and its derivative
pw = 1:30; % possible widths
ssq = sigma^2;
width = find(exp(-(pw.*pw)/(2*ssq))>GaussianDieOff,1,'last');
if isempty(width)
width = 1; % the user entered a really small sigma
end
t = (-width:width);
gau = exp(-(t.*t)/(2*ssq))/(2*pi*ssq); % the gaussian 1D filter
% Find the directional derivative of 2D Gaussian (along X-axis)
% Since the result is symmetric along X, we can get the derivative along
% Y-axis simply by transposing the result for X direction.
[x,y]=meshgrid(-width:width,-width:width);
dgau2D=-x.*exp(-(x.*x+y.*y)/(2*ssq))/(pi*ssq);
% Convolve the filters with the image in each direction
% The canny edge detector first requires convolution with
% 2D Gaussian, and then with the derivative of a Gaussian.
% Since Gaussian filter is separable, for smoothing, we can use
% two 1D convolutions in order to achieve the effect of convolving
% with 2D Gaussian. We convolve along rows and then columns.
%smooth the image out
aSmooth=imfilter(a,gau,'conv','replicate'); % run the filter across rows
aSmooth=imfilter(aSmooth,gau','conv','replicate'); % and then across columns
%apply directional derivatives
ax = imfilter(aSmooth, dgau2D, 'conv','replicate');
ay = imfilter(aSmooth, dgau2D', 'conv','replicate');
mag = sqrt((ax.*ax) + (ay.*ay));
magmax = max(mag(:));
if magmax>0
mag = mag / magmax; % normalize
end
% Select the thresholds
if isempty(thresh)
counts=imhist(mag, 64);
highThresh = find(cumsum(counts) > PercentOfPixelsNotEdges*m*n,...
1,'first') / 64;
lowThresh = ThresholdRatio*highThresh;
thresh = [lowThresh highThresh];
elseif length(thresh)==1
highThresh = thresh;
if thresh>=1
error(message('images:edge:singleThresholdOutOfRange'))
end
lowThresh = ThresholdRatio*thresh;
thresh = [lowThresh highThresh];
elseif length(thresh)==2
lowThresh = thresh(1);
highThresh = thresh(2);
if (lowThresh >= highThresh) || (highThresh >= 1)
error(message('images:edge:thresholdOutOfRange'))
end
end
% The next step is to do the non-maximum suppression. We will accrue
% indices which specify ON pixels in strong edgemap The array e will become
% the weak edge map.
e = cannyFindLocalMaxima(ax,ay,mag,lowThresh);
if ~isempty(e)
[rstrong,cstrong] = find(mag>highThresh & e);
if ~isempty(rstrong) % result is all zeros if idxStrong is empty
e = bwselect(e, cstrong, rstrong, 8);
e = bwmorph(e, 'thin', 1); % Thin double (or triple) pixel wide contours
end
end
elseif any(strcmp(method, 'log','zerocross'))
% The output edge map:
e = false(m,n);
rr = 2:m-1; cc=2:n-1;
% We don't use image blocks here
if isempty(H)
fsize = ceil(sigma*3) * 2 + 1; % choose an odd fsize > 6*sigma;
op = fspecial('log',fsize,sigma);
else
op = H;
end
op = op - sum(op(:))/numel(op); % make the op to sum to zero
b = imfilter(a,op,'replicate');
if isempty(thresh)
thresh = 0.75 * sum(abs(b(:)),'double') / numel(b);
end
% Look for the zero crossings: +-, -+ and their transposes
% We arbitrarily choose the edge to be the negative point
[rx,cx] = find( b(rr,cc) < 0 & b(rr,cc+1) > 0 ...
& abs( b(rr,cc)-b(rr,cc+1) ) > thresh ); % [- +]
e((rx+1) + cx*m) = 1;
[rx,cx] = find( b(rr,cc-1) > 0 & b(rr,cc) < 0 ...
& abs( b(rr,cc-1)-b(rr,cc) ) > thresh ); % [+ -]
e((rx+1) + cx*m) = 1;
[rx,cx] = find( b(rr,cc) < 0 & b(rr+1,cc) > 0 ...
& abs( b(rr,cc)-b(rr+1,cc) ) > thresh); % [- +]'
e((rx+1) + cx*m) = 1;
[rx,cx] = find( b(rr-1,cc) > 0 & b(rr,cc) < 0 ...
& abs( b(rr-1,cc)-b(rr,cc) ) > thresh); % [+ -]'
e((rx+1) + cx*m) = 1;
% Most likely this covers all of the cases. Just check to see if there
% are any points where the LoG was precisely zero:
[rz,cz] = find( b(rr,cc)==0 );
if ~isempty(rz)
% Look for the zero crossings: +0-, -0+ and their transposes
% The edge lies on the Zero point
zero = (rz+1) + cz*m; % Linear index for zero points
zz = (b(zero-1) < 0 & b(zero+1) > 0 ...
& abs( b(zero-1)-b(zero+1) ) > 2*thresh); % [- 0 +]'
e(zero(zz)) = 1;
zz = (b(zero-1) > 0 & b(zero+1) < 0 ...
& abs( b(zero-1)-b(zero+1) ) > 2*thresh); % [+ 0 -]'
e(zero(zz)) = 1;
zz = (b(zero-m) < 0 & b(zero+m) > 0 ...
& abs( b(zero-m)-b(zero+m) ) > 2*thresh); % [- 0 +]
e(zero(zz)) = 1;
zz = (b(zero-m) > 0 & b(zero+m) < 0 ...
& abs( b(zero-m)-b(zero+m) ) > 2*thresh); % [+ 0 -]
e(zero(zz)) = 1;
end
else % one of the easy methods (roberts,sobel,prewitt)
if isPrewittOrSobel
isSobel = strcmp(method, 'sobel');
scale = 4; % for calculating the automatic threshold
offset = [0 0 0 0]; % offsets used in the computation of the threshold
[bx, by, b] = edgesobelprewittmex(a, isSobel, kx, ky);
elseif strcmp(method, 'roberts')
x_mask = [1 0; 0 -1]/2; % Roberts approximation to diagonal derivative
y_mask = [0 1;-1 0]/2;
scale = 6;
offset = [-1 1 1 -1];
% compute the gradient in x and y direction
bx = imfilter(a,x_mask,'replicate');
by = imfilter(a,y_mask,'replicate');
% compute the magnitude
b = kx*bx.*bx + ky*by.*by;
else
error(message('images:edge:invalidEdgeDetectionMethod', method))
end
if (nargout > 2) % if gradients are requested
gv_45 = bx;
gh_135 = by;
end
% Determine the threshold; see page 514 of
% "Digital Imaging Processing" by William K. Pratt
if isempty(thresh) % Determine cutoff based on RMS estimate of noise
% Mean of the magnitude squared image is a
% value that's roughly proportional to SNR
cutoff = scale * sum(b(:),'double') / numel(b);
thresh = sqrt(cutoff);
else
% Use relative tolerance specified by the user
cutoff = (thresh).^2;
end
if thinning
e = computeedge(b,bx,by,kx,ky,int8(offset),100*eps,cutoff);
else
e = b > cutoff;
end
end
if nargout==0
imshow(e);
else
eout = e;
end
if isempty(a)
if nargout==2
if nargin == 2
if strcmp(method,'canny')
thresh = nan(1,2);
else
thresh = nan(1);
end
end
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : parse_inputs
%
function [I,Method,Thresh,Sigma,Thinning,H,kx,ky] = parse_inputs(varargin)
% OUTPUTS:
% I Image Data
% Method Edge detection method
% Thresh Threshold value
% Sigma standard deviation of Gaussian
% H Filter for Zero-crossing detection
% kx,ky From Directionality vector
narginchk(1,5)
I = varargin1;
validateattributes(I,'numeric','logical','real','nonsparse','2d',mfilename,'I',1);
% Defaults
Method = 'sobel';
Direction = 'both';
Thinning = true;
methods = 'canny','approxcanny','canny_old','prewitt','sobel','marr-hildreth','log','roberts','zerocross';
directions = 'both','horizontal','vertical';
options = 'thinning','nothinning';
% Now parse the nargin-1 remaining input arguments
% First get the strings - we do this because the interpretation of the
% rest of the arguments will depend on the method.
nonstr = []; % ordered indices of non-string arguments
for i = 2:nargin
if ischar(varargini)
str = lower(varargini);
j = find(strcmp(str,methods));
k = find(strcmp(str,directions));
l = find(strcmp(str,options));
if ~isempty(j)
Method = methodsj(1);
if strcmp(Method,'marr-hildreth')
error(message('images:removed:syntax','EDGE(I,''marr-hildreth'',...)','EDGE(I,''log'',...)'))
end
elseif ~isempty(k)
Direction = directionsk(1);
elseif ~isempty(l)
if strcmp(optionsl(1),'thinning')
Thinning = true;
else
Thinning = false;
end
else
error(message('images:edge:invalidInputString', varargin i ))
end
else
nonstr = [nonstr i]; %#ok<AGROW>
end
end
% Now get the rest of the arguments
[Thresh,Sigma,H,kx,ky] = images.internal.parseNonStringInputsEdge(varargin,Method,Direction,nonstr);
validateattributes(Thresh,'numeric','real',mfilename,'thresh',3);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : smoothGradient
%
function [GX, GY] = smoothGradient(I, sigma)
% Create an even-length 1-D separable Derivative of Gaussian filter
% Determine filter length
filterExtent = ceil(4*sigma);
x = -filterExtent:filterExtent;
% Create 1-D Gaussian Kernel
c = 1/(sqrt(2*pi)*sigma);
gaussKernel = c * exp(-(x.^2)/(2*sigma^2));
% Normalize to ensure kernel sums to one
gaussKernel = gaussKernel/sum(gaussKernel);
% Create 1-D Derivative of Gaussian Kernel
derivGaussKernel = gradient(gaussKernel);
% Normalize to ensure kernel sums to zero
negVals = derivGaussKernel < 0;
posVals = derivGaussKernel > 0;
derivGaussKernel(posVals) = derivGaussKernel(posVals)/sum(derivGaussKernel(posVals));
derivGaussKernel(negVals) = derivGaussKernel(negVals)/abs(sum(derivGaussKernel(negVals)));
% Compute smoothed numerical gradient of image I along x (horizontal)
% direction. GX corresponds to dG/dx, where G is the Gaussian Smoothed
% version of image I.
GX = imfilter(I, gaussKernel', 'conv', 'replicate');
GX = imfilter(GX, derivGaussKernel, 'conv', 'replicate');
% Compute smoothed numerical gradient of image I along y (vertical)
% direction. GY corresponds to dG/dy, where G is the Gaussian Smoothed
% version of image I.
GY = imfilter(I, gaussKernel, 'conv', 'replicate');
GY = imfilter(GY, derivGaussKernel', 'conv', 'replicate');
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : selectThresholds
%
function [lowThresh, highThresh] = selectThresholds(thresh, magGrad, PercentOfPixelsNotEdges, ThresholdRatio, ~)
[m,n] = size(magGrad);
% Select the thresholds
if isempty(thresh)
counts=imhist(magGrad, 64);
highThresh = find(cumsum(counts) > PercentOfPixelsNotEdges*m*n,...
1,'first') / 64;
lowThresh = ThresholdRatio*highThresh;
elseif length(thresh)==1
highThresh = thresh;
if thresh>=1
error(message('images:edge:singleThresholdOutOfRange'))
end
lowThresh = ThresholdRatio*thresh;
elseif length(thresh)==2
lowThresh = thresh(1);
highThresh = thresh(2);
if (lowThresh >= highThresh) || (highThresh >= 1)
error(message('images:edge:thresholdOutOfRange'))
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : thinAndThreshold
%
function H = thinAndThreshold(dx, dy, magGrad, lowThresh, highThresh)
% Perform Non-Maximum Suppression Thining and Hysteresis Thresholding of
% Edge Strength
% We will accrue indices which specify ON pixels in strong edgemap
% The array e will become the weak edge map.
E = cannyFindLocalMaxima(dx,dy,magGrad,lowThresh);
if ~isempty(E)
[rstrong,cstrong] = find(magGrad>highThresh & E);
if ~isempty(rstrong) % result is all zeros if idxStrong is empty
H = bwselect(E, cstrong, rstrong, 8);
else
H = false(size(E));
end
else
H = false(size(E));
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Local Function : computeapproxcanny
%
function e = computeapproxcanny(a, thresh)
a = im2uint8(a);
if isempty(a)
e = logical([]);
else
if isempty(thresh)
e = images.internal.ocvcanny(a);
else
if numel(thresh) == 1
e = images.internal.ocvcanny(a, 0.4*thresh, thresh);
else
e = images.internal.ocvcanny(a, thresh(2), thresh(1));
end
end
e = logical(e);
end
（可以在matlab中输入不带引号的“edit edge”来查看代码）
有没有人有任何建议或有与 matlab edge() 函数相同的精巧边缘检测器 Python 代码？任何建议或答案都会非常有帮助。谢谢。
【问题讨论】：
scikit-image.org/docs/dev/auto_examples/edges/…，这有用吗？ @xdze2 我尝试在 scikit 中使用 feature.canny，但它仍然会产生与 edge('canny') matlab 方法不同的结果。不过非常感谢您的回复 【参考方案1】：请说明你的意图：
你想用 python 写一个函数，它相当于 MATLAB 的 edge() 函数，并带有 Canny 实现吗？ （如果是这样，你为什么把所有选项都放在你最后的代码引用中？你只需要 strcmp(method,'canny') 部分）
如果您已经拥有如此出色的 MATLAB 代码来揭示精明的 impl。为什么不逐行翻译呢？
一般来说，要走的路是：
编写您自己的自定义 MATLAB 函数，该函数只执行您需要的功能，没有其他功能。如果您想要精明，请编写您自己的 MATLAB 函数来实现这一点。
在获得最少的工作 MATLAB 代码后，将其逐行转换为 python，同时在每一步之后从 MATLAB 代码和 python 代码输出调试图像，以验证您是否走在正确的轨道上。
最后提示：如果您发现使用 OpenCV 等效函数对您不起作用，您将不得不用 C++ 编写您的 OpenCV 代码，因为只有这样您才能使用调试器进入 OpenCV 的实现并找到了解它与 MATLAB 实现有何不同。
【讨论】：
感谢您的反馈，Shmuel Fine。 1.) 我想在 python 中编写一个函数，它等效于 MATLAB 的 edge() 函数，仅使用 Canny 实现。我只是忘记删除函数的 if 语句 2 的其余部分。）我正在制作一个需要在 Python 中进行精确边缘检测的程序。我在 matlab 中找到了我想在我的程序中实现的精巧边缘代码，但它使用了 edge() 函数（这是这行代码 mask = double(edge(mask,'canny',threshold));）。 你的问题中引用了边缘函数实现，它以 "function [eout,thresh,gv_45,gh_135] = edge(varargin)" 开头，不是吗？查看 varagin 解析： [a,method,thresh,sigma,thinning,H,kx,ky] = parse_inputs(args:);在你的情况下 **a**=mask, ** method**='canny'.... ?以上是关于如何在 Python 中像 edge() Matlab 函数一样制作精巧的边缘检测器的主要内容，如果未能解决你的问题，请参考以下文章
如何在 iOS 中像 UITabBarItem 一样设计 UIButton？