Allowing double type support if possible (Feature #2671)
Description
For 64 bit machine, which is the major type currently, operations on double is faster/equal to operations on float. Since the memory difference between double and float is not important for many people, I tried to use double in my code anywhere where float is used before so that don't care too much about type conversion.
A problem is that, some opencv functions only support float (CV_32F). I feel strange why don't them support for double (CV_64F). One example of this function is cvtColor.
Associated revisions
Merge pull request #2671 from syilma:HomographyDecomp
History
Updated by Alexander Smorkalov about 12 years ago
It is very rare feature for computer vision algorithm. But if you have any suggestions, you can contribute to OpenCV via Github pull request. See this page for more details: https://help.github.com/articles/using-pull-requests
- Category set to core
Updated by Ming-Ming Cheng almost 12 years ago
Thank you very much for the reply. I agree that most computer vision algorithms needs only float. I just feel that I can consistently use double since I typically don't care about the memory cost introduced by this. Consistently using double, like what was done in Matlab, avoids possible data type conversions when double is required. Modern 64bit computers process double at similar or even more efficiency than float.
Alexander Smorkalov wrote:
It is very rare feature for computer vision algorithm. But if you have any suggestions, you can contribute to OpenCV via Github pull request. See this page for more details: https://help.github.com/articles/using-pull-requests
Updated by Maksim Shabunin over 9 years ago
Issue has been transferred to GitHub: https://github.com/Itseez/opencv/issues/4477