Translation of serial camera data to RGB image.

Being a fairly new Matlab user, I have trouble translating the bytes I get back from a serial camera to a viewable, RGB565 image. The camera I'm using is this and so far I've managed to get 153600 bytes, back (320*240*2 for a 16 bit image) and store it in a 153600x1 column vector (not mentioning the grayscale image case with which I had no problem - what I did there was just reshape my 76800 bytes to a 320x240 matrix and then used imshow).

So my main question is how can I translate this vector to a proper matrix that I can view with imshow for example. I understand that an RGB image is an m-by-n-by-3 array but I can't seem to figure out how to reshape my vector as such.

Thank you very much in advance!

[M. Kiriakos]

For anyone interested, I managed to get the desired image with the help of a fellow from the uCam forum, suggesting something like this:

ind=1;
matrix_bin=dec2bin(frame,8); %frame is a 1x153600 vector of doubles

for j=1:2:153600
matrix_idx(ind)=bin2dec([matrix_bin(j,:) matrix_bin(j+1,:)]);                                                                                                                                                                                                                        
ind=ind+1;
end

im=reshape(matrix_idx,320,240);

imgR = uint8((255/31)*bitshift(bitand(frame_rot,63488),-11));  %# Red  
imgG = uint8((255/63)*bitshift(bitand(frame_rot,2016),-5));    %# Green 
imgB = uint8((255/31)*bitand(frame_rot,31));                   %# Blue 

my_img(:,:,1)=imgR;
my_img(:,:,2)=imgG;
my_img(:,:,3)=imgB;

imshow(my_img)

Thank you again for your time and suggestions!

[Walter Roberson]

The dec2bin() and bin2dec() can be avoided by using typecast()

[aseok]

Hello.

What is format of 'frame' vector ? my mean is how RGB565 data of raws and columns of a frame are provided in this vector? What is the frame_rot variable? Thanks.

[Walter Roberson]

aseok, see http://www.theimagingsource.com/en_US/support/documentation/icimagingcontrol-class/PixelformatRGB565.htm "Accessing the buffer"

Answer 1

[Walter Roberson]

If the two bytes of each 16 bit value occur one after the other, then use typecast() to convert the two bytes to a uint16 pair. After that you can reshape(), and possibly permute() if you need to move the planes around.

[M. Kiriakos]

Thank you for your response.

Well, when I try to typecast to uint16 and after manipulating the matrix a little, I manage to get an image that looks like this this, whereas I'm expecting something that looks like this. (The second image is taken directly from the camera software) The code I'm using is this:

uframe = uint8(frame4);  

Y = typecast(uframe, 'uint16');

frame_res_col = reshape((Y), 320, 240);

frame_flip_col = fliplr(frame_res_col);

frame_rot_col = imrotate(frame_flip_col, 90);

imshow(frame_rot_col)

frame4 is a 153600x1 vector of doubles and it's the vector I'm using to store the values read from the serial port. I then cast it to uint8 and then with typecast I try to read two bytes for each pixel. Judging from the image I'm getting, I'm not totally 'off' as this is the general picture I'm aiming for. But the colors are not correct. (The flipping and rotation needed to display the image properly, I discovered by trial and error)

So...any further suggestions would be more than welcome. Thanx again!

[Walter Roberson]

See http://www.imagingcontrol.com/en_US/support/documentation/class/PixelformatRGB565.htm

You will need to use bitget() and some shifting on the uint16 array in order to extract the color panes.

[M. Kiriakos]

I guess you mean trying something along those lines, right? I will definitely have to look into it as I am not that familiar with Matlab and coding in general. Do you think a modification of the above code (the one from Stackoverflow I mean) would work in my case?

[Walter Roberson]

The code there looks plausible.

Answer 2

[Image Analyst]

Well why did it end up as a 1D list of pixels? Didn't you use getsnapshot()?

[Walter Roberson]

The interface to the camera is via its serial port.

[Image Analyst]

Mine are too - universal serial bus. I didn't look up his/hers until now. My advice would be to dump that primitive OEM camera and get something easier to work with. Why spend all that time trying to get that thing to work when there are plenty of inexpensive cameras or webcams that will work instantly right out of the box? It's just not worth the time and trouble if you can't get it up and running in short order.

[M. Kiriakos]

I understand what you're saying and thank you for your response. I just think that the point is to solve the problem I am having with this camera. Not get another one because I couldn't solve the problem with this one... I wasn't aware of the getsnapshot() command to be honest but judging from Walter's post, I guess it doesn't work with a serial camera. Or am I mistaken? Unfortunately I will not be able to test it until tomorrow but nevertheless a serial object is indeed created for my camera with the serial command.

As for the 1D, I think this is just the way the camera outputs data. In a serial manner and then I just store them in a vector. As I mentioned above, I didn't have any problem manipulating the data in the grayscale case as it was just the intensity values. But the RGB case is a little trickier for me...

Answer 3

[aseok]

Hello. What is format of 'frame' vector ? my mean is how RGB565 data of raws and columns of a frame are provided in this vector? What is the frame_rot variable? Thanks.