How can I improve my segmented image?

Hi, I wrote the code below in order to segment blood vessels of the retina. But the result is not really good. Even I tried another way that I put it here to but it didn't make it better too. Can anyone give me a suggestion about how can I improve it? My images are from STARE database (<http://cecas.clemson.edu/~ahoover/stare/>) here is the example of my image:

FIRST WAY:

      clc; clear all; close all;

      %%Specifying path folders

      %%%ex.d_img2 - F:\folder1\folder2\...\foldern\*.format %%%
      %%%ex.di_img2 - F:\folder1\folder2\...\foldern\ %%%

      d_img2 = 'F:\Uni\Project\MATLAB\myMatlab\prepImages\*.ppm';
      di_img2 = 'F:\Uni\Project\MATLAB\myMatlab\prepImages\';

      %%Reading images

      srcFiles_prepImage = dir(d_img2);
      N_img_prep = length(srcFiles_prepImage);

      % i = 1;          %%50
      % i = 2;          %%150
      i = 3;          %%200

      filename2 = strcat(di_img2, srcFiles_prepImage(i).name);       
      prepImage = imread(filename2);
      prepImage = prepImage(:,:,2);         %?% the picture had been saved in rgb format?

      figure
      imshow(prepImage, 'InitialMagnification', 'fit')

      %%Histogram

      [row,column] = size(prepImage);

      figure
      imhist(prepImage);

      %%Creating & Applying binary mask in order to making the background (outside of eye) black

      %Using threshold
      threshold_value = 100;
      mask = prepImage > threshold_value; % Bright objects will be chosen.
      mask = imfill(mask, 'holes'); % Do a "hole fill" to get rid of any background pixels or "holes".
      %%%%check if there is any black part inside the white part
      figure
      imshow(mask, 'InitialMagnification', 'fit')

      % Applying binary mask to image
      maskedImage = bsxfun(@times, prepImage, cast(mask,class(prepImage)));
      figure
      imshow(maskedImage, 'InitialMagnification', 'fit')
      figure
      subplot(1,3,1)
      imhist(maskedImage)

      %%Thresholding histogram

      [pixelCount,grayLevels] = imhist(maskedImage,256);

      peaks = findpeaks(pixelCount);
      peaks = transpose(peaks);

      maxBin = max(peaks);
      maxBinLocation = find(pixelCount == maxBin);
      maxI = find(peaks == maxBin);

      % Find low threshold
      TlowLocation = min(find(pixelCount == 0));     % Find first zero
      % Place vertical bar on histogram to show Tlow
      subplot(1,3,3)
      imhist(maskedImage)
      hold on
      y2 = ylim();
      line([TlowLocation, TlowLocation], [y2(1), y2(2)], ...
        'Color', 'r', 'LineWidth', 2);

      % Find high threshold
      minBinL = min(peaks);   %%Find the minimum on left side of peak
      minBinLocationL = find(pixelCount == minBinL);
          %%%%
      % if minBinLocation>maxBinLocation
          %%%%need an IF to make sure that it is on left side
          %%%%
      % Find the minimum on left side of peak
      minIL = find(peaks == minBinL);
      leftBins = peaks(minIL:maxI);
      leftMean = mean(leftBins);
      subLeft = maxBin - leftMean;
      % Find the closest value to subLeft
      [subLeftClosest,indexLeftClosest] = min(abs(leftBins - subLeft));
      ThighBin = leftBins(indexLeftClosest);
      ThighLocation = find(pixelCount == ThighBin);
      % Place vertical bar on histogram to show Thigh
      subplot(1,3,2)
      imhist(maskedImage)
      hold on
      yl = ylim();
      line([ThighLocation, ThighLocation], [yl(1), yl(2)], ...
        'Color', 'r', 'LineWidth', 2);

      % Place vertical bars on histogram to show Tlow&Thigh
      figure
      imhist(maskedImage)
      hold on
      yl = ylim();
      line([TlowLocation, TlowLocation], [yl(1), yl(2)], ...
        'Color', 'r', 'LineWidth', 2);
      hold on
      y2 = ylim();
      line([ThighLocation, ThighLocation], [y2(1), y2(2)], ...
        'Color', 'r', 'LineWidth', 2);

      % Find pixels between the thresholds
      betweenThresholds = maskedImage<ThighLocation & maskedImage>TlowLocation;
      Vs = betweenThresholds;
      Ts = length(Vs);
      figure
      imshow(betweenThresholds, 'InitialMagnification', 'fit');

SECOND WAY:

      clc; clear all; close all;

      %%Specifying path folders

      %%%ex.d_img2 - F:\folder1\folder2\...\foldern\*.format %%%
      %%%ex.di_img2 - F:\folder1\folder2\...\foldern\ %%%

      d_img2 = 'F:\Uni\Project\MATLAB\myMatlab\prepImages\*.ppm';
      di_img2 = 'F:\Uni\Project\MATLAB\myMatlab\prepImages\';

      %%Reading images

      srcFiles_prepImage = dir(d_img2);
      N_img_prep = length(srcFiles_prepImage);

      % i = 1;          %%50
      % i = 2;          %%150
      i = 3;          %%200

      filename2 = strcat(di_img2, srcFiles_prepImage(i).name);       
      prepImage = imread(filename2);
      IG = uint8(prepImage*255);
      prepImage = prepImage(:,:,2);         %?% the picture had been saved in rgb format?

      figure
      imshow(prepImage, 'InitialMagnification', 'fit')

      %%Histogram

      [row,column] = size(prepImage);

      figure
      imhist(prepImage,256);

      %%Thresholding histogram

      [pixelCounts,grayLevels] = imhist(prepImage,256);

      cdf = cumsum(pixelCounts); % Sum histogram counts to get cumulative distribution function
      cdf = cdf / cdf(end); % Normalize
      % Get data value where ?%&?% is.
      data30 = find(cdf>= 0.2, 1, 'first');
      data70 = find(cdf>= 0.4, 1, 'first');

      % Place vertical bars on histogram to show Tlow&Thigh
      figure
      imhist(prepImage)
      hold on
      y1 = ylim();
      line([data30, data30], [y1(1), y1(2)], ...
        'Color', 'r', 'LineWidth', 2);
      hold on
      y2 = ylim();
      line([data70, data70], [y2(1), y2(2)], ...
        'Color', 'r', 'LineWidth', 2);

      % Find pixels between the thresholds
      betweenThresholds = prepImage<data70 & prepImage>30;
      Vs = betweenThresholds;
      Ts = length(Vs);
      figure
      imshow(betweenThresholds, 'InitialMagnification', 'fit');

      %%Creating & Applying binary mask in order to making the background (outside of eye) black

      %Using threshold
      threshold_value = 100;
      mask = prepImage > threshold_value; % Bright objects will be chosen.
      mask = imfill(mask, 'holes'); % Do a "hole fill" to get rid of any background pixels or "holes".
      %%%%check if there is any black part inside the white part
      figure
      imshow(mask, 'InitialMagnification', 'fit')
      % Applying binary mask to image
      maskedImage = bsxfun(@times, betweenThresholds, cast(mask,class(betweenThresholds)));
      figure
      imshow(maskedImage, 'InitialMagnification', 'fit')

Answer 1

[Image Analyst]

Basically you're just doing a primitive global histogram. No matter how you decide on a global threshold, it won't get all the vessels, no matter which threshold you pick.

You'd be better off using one of the successful published algorithms here: http://www.visionbib.com/bibliography/contentsmedical.html#Medical%20Applications,%20CAT,%20MRI,%20Ultrasound,%20Heart%20Models,%20Brain%20Models

[Ghazal Hnr]

Thank you for your response. The link seems really useful.