Hello, every one
1) I load the "barbara" image in MATLAB using the following command:
load woman
It is saved in workspace as "X" and its matrix class is "double". Then, I define "Y" as follows:
Y = uint8 (X);
the "Y" class is "uint8". The elements in "X" and "Y" are the same but when I use "imshow(X)" and "imshow(Y)", the results are different. why?
********************************************************************
2) When I use commands "[AX HX VX DX] = dwt2(X,'haar')" and "[AY HY VY DY] = dwt2(Y,'haar')" the results are the same but matrix elements of AX, HX, VX, DX are not in the interval [0 - 255]. Is it true?
********************************************************************
3) I use the following commands to save 3 types of the image in the current directory.
imwrite (X, 'barbara.bmp');
imwrite (X, 'barbara.tif');
imwrite (X, 'barbara.jpg');
Then I read them and save into 3 different matrices, X1, X2 and X3 as follows:
X1 = imread ('barbara.bmp');
X2 = imread ('barbara.tif');
X3 = imread ('barbara.jpg');
After that I see X1=X2 but X1 and X3 are different. why?
********************************************************************
4) I uses "dwt2" function for X, X1,X2 and X3. The results are different. why? Which format of the image should be used for DWT?
I look forward to hearing from researchers who know the answers. Thank you.
>> load woman
>> Y = uint8 (X);
>> whos X Y
Name Size Bytes Class Attributes
X 256x256 524288 double
Y 256x256 65536 uint8
the type of X and Y are different
Answer 1 : This is essentially an indexed image, so there is also a colormap called "map" loaded in memory. The proper way to display an indexed image is : imshow (X, map). Moreover, since the datatype is double (64-bit), MATLAB expects all values to lie between 0 and 1. Values more than 1 are displayed as white, therefore only a white square is displayed if you simply use imshow(X). For confirmation, if you open the file "map" you will see 3 columns representing R,G,B values and all values range from 0 to 1 (since this is a grayscale image, values in 3 columns will be equal). Alternatively, you can also map the range [0, 1] to [0, 255] by using the command : imshow(X, [ ]). Since the original datatype was 'double' values were represented using 64-bit precision. When this is converted to uint8 , precision becomes only 8-bit, so some values might be lost, resulting in a difference between image appearances.
Answers to remaining questions will follow.
Hello, every one.
Thank you so much for your responses.
Best Regards
4. If you have a color image, you can't uses dtw2. Dwt2 Function works for grayscale image
Answer 2 : The coefficients returned by dwt2 all have data type 'double' so obviously their values would not be in the range [0, 255] which is true only for 'uint8'
Answer 3 : Since X is an indexed image with a colormap 'map' the syntax for saving needs to be : imwrite(X, map, filename). Also since JPG is a lossy compression format, the results would be different from BMP and TIFF which are lossless formats.
Answer 4 : Any format which is either non-compressed or implements a non-lossy compression is preferable, as lossy compression degrades the image by deleting portions of the data. Hope all the queries are answered satisfactorily.
Dear Dr. Parekh
Many thanks for your really complete responses.
King Regards
- Badges
- Science topic