Niblack算法实现Matlab

Question

我在Matlab中实现了Niblack的算法。但是，我正在取得一些可疑的结果。

我下载了一些试图执行某些文档二值化的测试图像。但是，我不确定它们看起来是否正确。下面是图像和我的代码，我试图通过执行各种功能使其模块化。谢谢！

将k级提升至2级 结果：

我的代码

% K value
k_threshold = -0.1;

%input image
image1 = imread('test1.jpg');

%convert to grayscale
image1 = rgb2gray(image1);

%convert to double
image1 = double(image1);

% normalise to [0,1] range
image1 = image1 / max(image1(:));

%mean filtering
mean = averagefilter2(image1);

%standard deviation
standardDeviation = standardDev(image1, mean);

%calculate binary image
binaryImage = image1 >=  (mean + k_threshold * standardDeviation);


%display image
display = displayImage(binaryImage, image1);

function deviation = standardDev(image1, mean)
    meanSquare = averagefilter2(image1.^2);
    deviation = (meanSquare - mean.^2).^0.5;

end
function [X,Y] = displayImage (binaryImage, image1)
close;

figure;
X = imshow(image1); title('original image');

figure;
 Y = imshow(binaryImage); title('binary image');

end

function img=averagefilter2(image1)
    meanFilter = fspecial('average',[3 3]);
    img = imfilter (image1,meanFilter);
end