Level the blobs and Delete Blob containing Text in it.

I want to delete texts or words present in my result image.

let the image is :

i have done the blob analysys and plot the object location using this code:

%% Calculate number of object present in output image

    BW=bwlabel(BW);

    s = regionprops(BW,'BoundingBox');

    b = bwboundaries(BW);

    n = size(b, 1);

%% Function For Ploting object location

 figure,imshow(BW)

 hold on;

 for n=1:size(s,1)

rectangle('Position',s(n).BoundingBox,'EdgeColor','y','LineWidth',1)

 end

 hold off;

so then the output image is :

now i want to lable these blobs:

and then remove the blob containing text in it.:

so thus the output image will be:

and the final image i want is :

How will i do this?

[Walter Roberson]

Why isn't bwlabel separating out the letters of the text into individual objects? The 'S' does not appear to be touching the 'K' at all; 'K' just might be touching the left tip of the 'Y' though.

Are you possibly doing a morphological filling operation before doing blob boundary detection?

Answer 1

[Image Analyst]

Priyanka, what's the status of this? Do you still need help? Can you answer Walter's question? Can you just filter based on the area of the blob? Do all the bird shapes have about the same area, which is different than the letters? Can you use this example? http://www.mathworks.com/help/vision/examples/automatically-detect-and-recognize-text-in-natural-images.html?prodcode=VP&language=en to find the text and then delete it?

[Priyanka Roy]

All birds shapes are not same. And the text may change time to time when image changes. the below code i use for creating the blob boundary based on the area:

hold on;

b = bwboundaries(K);

numberOfBoundaries = size(b, 1);

for k = 1 : numberOfBoundaries

thisBoundary = b{k};

plot(thisBoundary(:,2), thisBoundary(:,1), 'g', 'LineWidth', 2);

end

hold off;

i want to remove the blob which containing the text.

[Image Analyst]

You marked it as accepted so I presume the "finding text in natural scenery" link is what you're using.