How does the winvideo adaptor in Image Acquisition Toolbox exactly return YUY2 data in YCbCr colorspace?

5 visualizaciones (últimos 30 días)
I have a camera (Leopard USB camera LI-USB30-M021X) with winvideo compatible DirectShow driver which says it outputs images in YUY2 format. Now YUY2 is a format in which you normally use 16-bit per pixel to describe a color image. This camera does not actually capture color images though, in reality it captures 12-bit grayscale images which are then encoded into those 16-bits per pixel YUY2 format. Can I recover the original 12-bit data format in MATLAB somehow?

Respuesta aceptada

MathWorks Support Team
MathWorks Support Team el 14 de Dic. de 2022
Editada: MathWorks Support Team el 26 de En. de 2023
If you look at how YUY2 data is stored:
 
We see, that for the first 4 pixels, it looks like:
Where with the colors I have tried to indicate how I assume the camera vendor has encoded the 12-bit data as 2-bytes of data in YUY2.
 
If using videoinput with a YUY2 format and the default YCbCr ReturnedColorSpace, MATLAB then maps this data to a YCbCr image "im" in MATLAB in the following way:
Y0 = im(1,1,1)
U0 = im(1,1,2) = im(1,2,2)
Y1 = im(1,2,1)
V0 = im(1,1,3) = im(1,2,3)
Y2 = im(1,3,1)
U1 = im(1,3,2) = im(1,4,2)
Y3 = im(1,4,1)
V1 = im(1,3,3) = im(1,4,3)
Since there is no scaling or averaging going on, you should be able to access the original raw 12-bit data by combining the right elements again. i.e. to get Y0|U0 we need to take im(1,1,1) and combine it with im(1,1,2) where, depending on how the vendor encoded the data exactly in which order, one of those two should be shifted by 8 bits (= multiplying by 256). So then to get the first 12-bit value you would use:
v = uint16(ImageFrame(1,1,2))*2^8 + uint16(ImageFrame(1,1,1));
Similarly for Y1|V0 we would combine im(1,2,1) with im(1,2,3). Or for a whole 1280x720 image we should then be able to write:
A = zeros(720,1280,'uint16');
A(:,1:2:end) = uint16(ImageFrame(:,1:2:end,2))*2^8 + uint16(ImageFrame(:,1:2:end,1));
A(:,2:2:end) = uint16(ImageFrame(:,2:2:end,3))*2^8 + uint16(ImageFrame(:,2:2:end,1));
 

Más respuestas (0)

Etiquetas

Aún no se han introducido etiquetas.

Productos


Versión

R2017b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by