There are some .tif format datasets about remote sensing. you know,channels of remote sensing are not least 3 or 4. I had tried IMREAD.UNCHANGED and I don't konw whether I chosed wrong flag or CV2 does not support this kind of channels
this suggests that 512 is the upper limit for cv::Mat: https://docs.opencv.org/3.4/d1/d1b/group__core__hal__interface.html#ga3de14a42631396fe0480be69d5d2363f
cv::imread uses third party libraries. they may or may not support more than four channels. opencv may not be prepared to accept more channels even if the library can do it.
opencv's imread is a simple function. instead you might want to use an appropriate library directly. libtiff may be suitable.
Asked: 2020-04-01 04:10:57 -0600
Seen: 1,516 times
Last updated: Apr 02 '20
AFAIK OpenCV imread doesn't support multichannel TIFF files. Here's my answer for reading them using libtiff directly.
okay.Let'm explore lib,but I wonder whether the zip module will zip it after the lib read tif.And I want to feed the mutichannerl TIFF format to model
if the tiff is <=4 channel, simple imread can load it. GDAL is also a fair option to load multispectral tiff images, by my knowledge it has a nice Python interface (C++ is a bit rougher tho), and theorically, if you build opencv with GDAL, the IMREAD_LOAD_GDAL flag should let you load multispectral images with simple imread. BUT I have concerns if it reads only 3 channels, and I wrote a function using the GDAL interface directly to load images. Libtiff probably better for C++ implementations.