We get features from sift algorithm and then match by ransac method. is it complex and repeat?
How about this processing if we get the points?
In an image, we select three points to combine a triangle, calculate two internal angles, then find three points to combine a triangle in another image, to see if they are similar by comparing their angles.
Is this simple and fast to get H matrix and match? I think it is.
Hi Dear
I found a useful project about this , please follow the link :
http://www.codeproject.com/Articles/95453/Automatic-Image-Stitching-with-Accord-NET
thanks
Dear Jek:
No it is not complex to combine SIFT & RANSAC. SIFT is based on the matching of image descriptors and its general computational complexity is O(n2). Yes indeed, some operations should be performed to account for the computation of the image descriptors and this is in addition to the pairing process that will be performed during the search process to build hypothesis for the matching candidates. On the other hand, RANSAC is based on the mathematical model that describes the geometrical relationship between two images or more (e.g., fundamental matrix). In summary, it is not difficult to build an iterative algorithm for image matching that combines SIFT & RANSAC. In fact, the RANSAC can act as a hypothesis verification mechanism for the matched pairs from the SIFT process.
Regards,
Gamal
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing