DTYPE not properly parsed [NotImplemented] #44

Activity

1. rbnvrw

   added this to the Version 3.3.1 milestone on Nov 10, 2021
2. rbnvrw commented on Nov 10, 2021

   

   rbnvrw

   on Nov 10, 2021

   Member

   More actions

   Thank you for your report @ggirelli and sorry for the late reply.
   Do you have a suggestion on how to implement this? Otherwise I'll decide what to do with it before the next release (either try to decide from the metadata or leave it as-is).
3. ggirelli commented on Nov 11, 2021

   

   ggirelli

   on Nov 11, 2021

   Last edited by ggirelli

   ContributorAuthor

   More actions

   So, the Parser._raw_metadata.SLxImageAttributes.ePixelType tag seems to match the SampleFormat TIFFTag. Hence, possible values should be:

   1 = unsigned integer data
   2 = two's complement signed integer data
   3 = IEEE floating point data
   4 = undefined data format
   

   We can combine this with the bits per pixel from the Parser.SLxImageAttributes.uiBpcInMemory tag to build a dtype compatible with numpy. How about something like:

   def propose_dtype(parser) -> str:
       bits_per_pixel = parser._raw_metadata.image_attributes[
               six.b("SLxImageAttributes")][six.b("uiBpcInMemory")]
       pixel_type_tag = parser._raw_metadata.image_attributes[
               six.b("SLxImageAttributes")][six.b("ePixelType")]
   
       dtype_base: DefaultDict = defaultdict(lambda: "float")
       dtype_base[1] = "uint"
       dtype_base[2] = "int"
   
       proposed_dtype = f"{dtype_base[pixel_type_tag]}{bits_per_pixel}"
   
       supported_dtypes = ("uint8", "uint16", "uint32",
                           "int8", "int16", "int32",
                           "float8", "float16", "float32")
       return "float64" if proposed_dtype not in supported_dtypes else proposed_dtype

   Hope this makes sense 😄