16-bit grayscale TIFF

What is the bit-format of a 16-bit grayscale TIFF? I read that only 10 bits actually contai开发者_开发百科n intensity data--is this true? When I try to open a 16-bit TIFF in Matlab using imshow(), the image is totally black, but the same image reduced to 8-bits displays fine.

How many bits are used depends a lot on the camera that took the picture. A 14-bit camera will use 14 of 16 bits, a 10-bit camera will use 10.

If you display the image by calling imshow(img), all will be black if the full dynamic range of the image wasn't used, since imshow will scale to the dynamic range (i.e. 0 to 2^16-1). If you call imshow(img,[]) instead, which scales the image to min/max, the image displays nicely.

If a TIFF creator is writing a well-behaved TIFF, each channel should be normalized to the full dynamic range of that channel. In other words, 16 bits per channel should range from 0-65535, 8 bits per channel should range from 0-255.

While the number of bits used could be encoded into a tag, there are no baseline tags that encode this. This means that if a 10 bit device doesn't normalize to 16 bits, there is no way for a baseline reader to understand the intent of the image.

As a side note, there are other image file formats (Dicom is one) which provide a way to express how many bits per channel are significant.