我如何将16位深度的图像（png格式）转换为32位的图像？

Question

I have a depth image 16 uint of an human hand, after thresholding. I want to recover some geometric features as contour, convex hull, etc., but all these functions work with 8 or 32 bit images so I need to convert. I'm working in python!

I've tried with some open cv function as

img32 = np.array(thresh, dtype=np.float32) # This line only change the type, not values
img32 *= 65536 # Now we get the good values in 32 bit format
print img.dtype

but I've always a 16 uint image.

Answer 1

I don't believe PNG supports floating point data, so I am not sure what format you expect the output to be in?

If raw 32-bit float RGB is ok, you could use ImageMagick which is installed on most Linux distros and readily avilable for OS X and Windows. The command would be

convert input.png -depth 32 -define quantum:format=floating-point rgb:out.rgb

For example, if you create a 1x1 white image like this

convert -size 1x1 xc:white -depth 32 -define quantum:format=floating-point rgb:out.rgb

you get a 12-byte image like this

xxd out.rgb
0000000: 0000 803f 0000 803f 0000 803f

If you create a 1x1 black image like this

convert -size 1x1 xc:black -depth 32 -define quantum:format=floating-point rgb:out.rgb

you get a 12-byte image like this

xxd out.rgb
0000000: 0000 0000 0000 0000 0000 0000