matlab code How to dived 400*500 image into 40*50 and calculate object area in each block. put all blocks in one imafe

I am trying to classify weed and crop. I did feature extraction.finally I got Binary image of the crop and weed.The last part isclassification based on area of weed and crop.

I am doing classification based on pixel area of binary image. For that I want divide wholw image into 40*50 of 100 blocks. Find the area in each block.If pixel B(i,j)=1, count= cont +1. If white pixel count in the block more 80% that is crop otherwise it is weed. For that I want process all blocks and pur result in one image.I have problem with looping plese help me. to do all.

Answer 1

[Turlough Hughes]

You can find out which blocks are crops (according to your definitions) as follows:

B = blockproc(Img, [40 50], @(block_struct)... 
(nnz(block_struct.data)/numel(block_struct.data))>0.8 )

B is just a 10x10 array corresponding to your blocks.

By wanting the results in one output image do you mean you want the blocks less than or equal to 80% to appear black? Try the following script.

clear all
close all
clc
fontSize = 16;

% Generate image
Img = false(400,500);
ind = randperm(numel(Img),numel(Img)*0.8);
Img(ind) = true; 

figure('units','normalized','outerposition',[0 0 1 1])
subplot(1,2,1);
imshow(Img)
title('Input image (approx. 80% white pixels).','fontSize',fontSize);

% Block process sets blocks which consist of <= 80% true pixels to all 0.
B = blockproc(Img, [40 50], @(block_struct) myfun(block_struct.data));

subplot(1,2,2);
imshow(B)
title('Block processing result','fontSize',fontSize)


function blockout = myfun(blockin)
    if nnz(blockin)/numel(blockin) > 0.8
        blockout = blockin;
    else
        blockout = false(size(blockin));
    end
end

[Turlough Hughes]

Have you tried this?

[Vinitha Patil]

thank you for yor help ,yes I tried , I am differentiating weed and crop from the image. I converted image into gray used global threshold using graythesh.

close all

I = imread('crop4.jpg');

figure,imshow(I);

[N,M,Z]= size(I);

I(:,:,[1 3])=0;

figure,imshow(I);

Im = rgb2gray(I);

imshow(Im);

figure,imhist(Im);

T2 = graythresh(Im);

It2 = im2bw(Im,T2);

figure,imshow(It2)

%to fill connected hole

Bw2 =imfill(It2,[3 3],8);

figure,imshow(Bw2)

title('hole');

[n,m]= size(Bw2);

J = imresize(Bw2, [400 500]);

figure,imshow(J);

Now i want to differentiate weed and crop based on features. Mark the weed .For that I tried to split the image into 100 blocks process each block by feature. Heare crop leaves are bigger in size and weed leaves smaller.

please help me to classify and display marked weed on image with code.

[Vinitha Patil]

Please help me out to solve the problem

[Turlough Hughes]

I don't think block processing will work.

I've played around with a few options and I think you could be going down a rabbit hole trying to segment that part of your image, having said that I am not an expert on segmenting images. The main challenge for me is inluding the outer rows of crops while removing the inner row of weeds at the same time. Anyway, I'll share the code I got so far which may or may not help...

% Intial Image provided by you (small adjustments)
I = imread('crop4.jpg');
[N,M,Z]= size(I);
I(:,:,[1 3])=0;
Im = rgb2gray(I);
BW = imbinarize(Im);
[n,m]= size(BW); 
Img = BW;
Img = imresize(Img,[400 500]);


fontSize = 16;
% Plot results
figure('units','normalized','outerposition',[0 0 1 1])
subplot(2,2,1), imshow(Img)
title('Input image for block processing','fontSize',fontSize);

% Block process 4 by 4 blocks and keep blocks w/ > 50%
B = blockproc(Img, [4 4], @(block_struct) blockfun(block_struct.data));
% Plot results
subplot(2,2,2), imshow(B)
title('After block processing','fontSize',fontSize)

% Remove connected areas <500 pixels.
thresh = 500; 
C = bwareaopen(B,thresh);
% Plot results
subplot(2,2,3), imshow(C)
title(sprintf('After bwareaopen (areas must be > %dpx)',thresh),'fontSize',fontSize)

D = imerode(C,strel('disk',2));
subplot(2,2,4), imshow(D);
title('Further imerode','fontSize',fontSize)
% function that is implemented on each block.
function blockout = blockfun(blockin)
    if nnz(blockin)/numel(blockin) > 0.5 % 50 percent of pixels must be white
        blockout = blockin;
    else
        blockout = false(size(blockin)); % return all black pixels if condition is not satisfied.
    end
end

Maybe someone else on the forum can come up with another way to segment your image, but I don't think the blockprocessing is particularly helpful.

Combinations of imerode and skeletonisation using the bwmorph function might also be worth trying.

[Vinitha Patil]

Thank you some how it will help

[Vinitha Patil]

Is it possible to find area of objects in each divided block of image. If area is in between 10 to 200 pixel then the object is weed mark the position or centroide of the weed on original plant image. Does it really work with blockproc() function. please help me out.

[Turlough Hughes]

If area is in between 10 to 200 pixel then the object is weed.

I tried and that doesn't distinuish your crops from your weeds, only one 40 by 50 block passes that condition. If you want to try different blockprocessing strategies then modify the local function at the bottom of the script I provided. If you have any question about how the above code works I am happy to explain further.

Answer 2

[Matt J]

Using sepblockfun from the File Exchange

https://www.mathworks.com/matlabcentral/fileexchange/48089-separable-block-wise-operations

Areas=sepblockfun(B,[40,50],'sum');