On 7-1-2019 11:45, clipka wrote:
> Am 07.01.2019 um 08:49 schrieb Thomas de Groot:
>> On 7-1-2019 8:40, Thomas de Groot wrote:
>>>
>>> Interesting. I missed that too. Going to play right now as I am 
>>> working on an isosurface...
>>>
>>
>> Hmmm... right. Not what I expected. Polarity may be only relevant if 
>> the contained_by shape is close to the isosurface one, i.e. spheres, 
>> cylinders, cubes. Otherwise, I don't understand its usefulness. 
>> Polarity on just renders the contained_by shape, whatever its 
>> (positive) value.
>>
>> Did I miss something? Probably.
> 
> The magnitude of the `polarity` paramezer is irrelevant, only the sign 
> matters (or, more precisely, whether the parameter is positive; zero has 
> the same effect as a negative value).
> 
> `polarity 1` should have the same effect as flipping the signs of both 
> the function and the threshold.
> 
> If your isosurface is fully inside the `contained_by` shpe, then it is 
> perfectly normal that `polarity 1` will cause you to see only the 
> `contained_by` shape, because everything outside the `contained_by` 
> shape is always considered "outside" (*), while inside that shape the 
> "inside" and "outside" are now the other way round.
> 
> To just flip "inside" and "outside" (including the space outside the 
> `contained_by` shape, you should use the `inverse` keyword instead.
> 
> 
> The `polarity` keyword is primarily intended to complement the 
> `potential` pattern feature, which would behave inconsistently between 
> blobs and isosurfaces unless positive polarity mode is used for the latter.

OK, I guess I understand. I then wonder if the polarity use by Jr to 
solve COMPATT's problem is correct. My uneducated guess would be 'no'. ;-)

-- 
Thomas

Post a reply to this message