OpenCV: cvLoadImage opens 16-bit image as 8-bit

I am trying to process 16-bit single-channel uncompressed TIFF images with OpenCV 2.1, but when I call cvLoadImage, apparently they 开发者_如何学Cget converted to 8-bit:

IplImage* img = cvLoadImage("myImage.tif",
      CV_LOAD_IMAGE_ANYDEPTH | CV_LOAD_IMAGE_ANYCOLOR);

std::cout << img->depth << std::endl;

prints

8

The images have been produced by a fluorescence scanner that generates only this kind of images, and I have confirmed with a commercial software package that they really are 16-bit.

How can I open these images and work with the original bit depth?

In OpenCV Loading a tif image (16 bits for example) in python should work by typing:

im = cv.LoadImageM("test.tif",cv.CV_LOAD_IMAGE_UNCHANGED)

I am using OpenCV 2.3.1, this version allows opening images of 16 bits like jpromvi made:

IplImage* img = cvLoadImage("myImage.tif",
      CV_LOAD_IMAGE_ANYDEPTH | CV_LOAD_IMAGE_ANYCOLOR);

Hmm... you could write your own code to load the image. Its uncompressed TIFF, so shouldn't be much trouble.

I had a similar problem (using OpenCV 2.x on Linux, with the latest libtiff). I was actually using 12bit tiff Images, which does not make it easier...

I can recommend ImageMagic.

You can verify that your image is really 16bit

display myImage.tiff (right click, "Image Info")

I converted it to png - which worked perfectly in my case

convert myImage.tiff myImage.png

use GDAL lib to open 16 bit tiff image, and then convert gdal dataset to opencv Mat

see: https://github.com/marvins/Code_Sandbox/blob/master/c%2B%2B/opencv/geo_tools/gdalToOpenCV.cpp