Reduce the image size while maintaining the quality. Use dct2 with RGB image. when I Concatenate the RGB I get white image

% import image

pic=imread('image3.jpg');

size_a=size(pic)

imshow(pic);

% The image size is 2448x3264x3

% split RGB channels

[R,G,B]=imsplit(pic);

% apply DCT to each channel

DCTR=dct2(R);

DCTG=dct2(G);

DCTB=dct2(B);

% obtain the upper left corner of DCT to keep important values

[n1,n2]=size(DCTR)

picR=DCTR(1:round(n1/2),1:round(n2/2));

picG=DCTG(1:round(n1/2),1:round(n2/2));

picB=DCTG(1:round(n1/2),1:round(n2/2));

% perform inverse of DCT on all channels

R1=idct2(picR);

G1=idct2(picG);

B1=idct2(picB);

% combine RGB channels

rgbImage=cat(3,R1,G1,B1)

imshow(rgbImage,[])

size(rgbImage)

Answer 1

[Mahesh Taparia]

Hi

The image get converted to double format, convert back to uint8 format to show correctly. For example:

imshow(uint8(rgbImage));

Answer 2

[Image Analyst]

Unlike gray scale images, you can't use [] with RGB images at all (it will just ignore them if you do). Also, if the RGB values are doubles and in the range 0-1, it will work. If they're outside that range you'll either have to scale it to that range with rescale() or mat2gray(), OR convert to uint8 in the range 0-255. Here's a little demo of some of the possibilities:

% Make double image in the range 0-3000
grayImage = 3000 * rand(100, 100, 3);
imshow(grayImage, []); % Works

% Make image in the range 0-255
rgbImage = 250 * rand(100, 100, 3);
imshow(rgbImage, []); % Doesn't work

% Make double image in the range 0-5000 and convert to uint8
rgbImage = uint8(255 * mat2gray(5000 * rand(100, 100, 3)));
imshow(rgbImage); % Works

% Make image in the range 0-65535 and convert to uint16
rgbImage = uint16(65535 * rand(100, 100, 3));
imshow(rgbImage); % Doesn't work

% Make image in the range 0-5000 and convert to uint8
rgbImage = im2uint8(5000 * rand(100, 100, 3));
imshow(rgbImage); % Doesn't work

% Make image in the range 0-1
rgbImage = 0.95 * rand(100, 100, 3);
imshow(rgbImage, []); % works