识别图像中不连续出现的最暗颜色

Question

I am looking for a way to identify the dark grey ovals on this example image:

I know I can find the darkest color in the image with a simple min . However, I am not interested in identifying the black in this example (darkest, contiguous part of the image). Therefore, I think my problem can be described: identifying the darkest color in the image that occurs noncontigously . That is, the dark grey ovals in the example above. I would appreciate your help. Notice how there's some noise also in the image, that I would ideally not take into account when find this darkest noncontiguous color. However, that might be different step/too much for this question.

I've been messing with output from regionprops to achieve this goal. However, I am not sure how to combine that output with color data. I was wondering if there is some MATLAB tool, or some image processing trick that would make this task easier.

Thanks in adavnce!

Answer 1

If the expected range/tones of grey to be extracted is known image thresholding can be applied to eliminate all the bodies that are not of that unique colour/tone. Here I converted the image to greyscale and filtered out the noise in the image using a median filter.

Image 1: Lower_Threshold : 125 and Upper_Threshold : 240

Image 2: Lower_Threshold : 20 and Upper_Threshold :160

  File_Name = "Ovals.png";

    Image = imread(File_Name);
    Greyscale_Image = rgb2gray(Image);
    
    %Filtering out some noise%
    Filtered_Image = medfilt2(Greyscale_Image, [10 10]);
    
    [Image_Height,Image_Width,Depth] = size(Image);
    
    Lower_Threshold = 125;
    Upper_Threshold = 140;
    
    
    %Thresholding the signal%
    for Row_Scanner = 1: +1: Image_Height
       for Column_Scanner = 1: +1: Image_Width 
        
        Pixel_Value = Filtered_Image(Row_Scanner,Column_Scanner);
        if(Pixel_Value > Upper_Threshold)
        Filtered_Image(Row_Scanner,Column_Scanner) = 255;    
        end
        
        if(Pixel_Value < Lower_Threshold)
            Filtered_Image(Row_Scanner,Column_Scanner) = 255;    
        end
        
       end
    end
    
    Filtered_Image = medfilt2(Filtered_Image, [10 10]);
    imshow(Filtered_Image);

Additional Sobel Edge Detection:

Lowering the Upper_Threshold can filter out more data that is not of interest.

File_Name = "PCA_Image.png";
Image = imread(File_Name);
Greyscale_Image = Image;

[Image_Height,Image_Width,Depth] = size(Image);

Lower_Threshold = 0;
Upper_Threshold = 50;

%Thresholding the sig6al%
for Row_Scanner = 1: +1: Image_Height
   for Column_Scanner = 1: +1: Image_Width 
    
    Pixel_Value = Greyscale_Image(Row_Scanner,Column_Scanner);
    if(Pixel_Value > Upper_Threshold)
    Greyscale_Image(Row_Scanner,Column_Scanner) = 255;    
    end
    
    if(Pixel_Value < Lower_Threshold)
        Greyscale_Image(Row_Scanner,Column_Scanner) = 255;    
    end
    
   end
end

Binary_Image = Greyscale_Image < 255;
Edge_Image = edge(Greyscale_Image,'sobel');


subplot(1,2,1); imshow(Binary_Image); 
subplot(1,2,2); imshow(Edge_Image);

Answer 2

These are steps you need to take to extract darker ovals from the image you provided:

1. Import the image
2. Turn it to black and white so you call regionprops and extract shape properties
3. Select the regions by their properties
4. Find the darkest color from selected regions

clc; close all; clear variables;
% import the image
I = im2double(rgb2gray(imread('KtzM7.png')));

% convert the image to balck and white
BW = (1-I)>0.05;

% extract shape and intensity properties of 
% connected components
stats = regionprops(BW, I, 'MinorAxisLength', ...
    'MajorAxisLength', 'Area', 'MeanIntensity');
% filter components by theis extend and area
ratio = [stats.MinorAxisLength] ./ [stats.MajorAxisLength];
[~, circleLike] = find((ratio > 0.65) & ([stats.Area] > 30));

% extract regions of interest in the image
L = bwlabel(BW);
RoI = ismember(L, circleLike);

% find the lowest intensity among RoI
intensities = [stats.MeanIntensity];
darkest = min(intensities(circleLike));

% remove non-RoI regions from output
O = max(I, 1-RoI);

% remove components that their intensity is not 
% equal to darkest color
O(abs(O-darkest)>0.1) = 1;

subplot 221, imshow(I), title('Input Image')
subplot 222, imshow(BW), title('Black & White')
subplot 223, imshow(RoI), title('Region of Interest');
subplot 224, imshow(O), title('Output Image')