Reconstruct the initial image with idct2

Hello, for an assignment I needed to create an image of 512x512 pixels black with in its center a square of white pixel of 100pixels, then I had to make its transformation into discrete cosine and now I must thanks to the function idct2, reconstruct the initial image by retaining only :

• The 80% largest coefficients of the DCT
• The 50% largest coefficients of the DCT
• The 20% largest coefficients of the DCT

But I don't know how to set this parameter, here is my code currently.

image=zeros(512,512);
image(206:305,206:305)=255;

J = dct2(image);
imshow(log(abs(J)),[])
colormap parula
colorbar

Answer 1

[Jonas]

have a look into this post

[Damien Dubois]

No I hadn't found this post. Thanks a lot for your help. Here is how I proceeded I think I have the solution to keep for example the 75% largest coefficient of my DCT. Can you confirm that this is the case?

image=zeros(512,512);
image(206:305,206:305)=255;

J = dct2(image);
Y=prctile(abs(J),75,'all');
J(abs(J)<=Y)=0;
imshow(rescale(idct2(J)))

[Jonas]

your code is correct. it will remove all values that are smaller than 75% border of absolute values. you can change it to 80, 50 (which is the median by the way!) and 20 to achieve your original goal according to your question

[Damien Dubois]

Thank you very much for your precious help