I know scale invariant SIFT feature but this technique is tedious. Another features are Histogram of Oriented Gradient HOG, which is efficient and can be rotation invariant. But I didn't find any feature which is invariant to RTI and scale changes and yet it's computation cost is low?
Can anyone please suggest me such features if it exists?
Can I use combination of these feature? if yes, how can I combine them as they have different size?
Complex Zernike moments are translation and rotation invariant and may be useful.
http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/SHUTLER3/node11.html
Hi,
Try Zernike moments, some related links are attached plz check
http://www.sciencedomain.org/abstract.php?iid=550&id=6&aid=4884
http://optics.sgu.ru/~ulianov/Bazarova/LASCA_literature/InvariantImageRecognitionZernikeMoments.pdf
http://www.csie.ntnu.edu.tw/~bbailey/Moments%20in%20IP.htm
Hi,
Hu moments are translation, scale, mirroring, and rotation invariant.
To compute SIFT features for images, binaries of these are available as VLFeat (open source package) at http://www.vlfeat.org/ with matlab and python codes.
Such library is written in C and also it has a command line interface that can be used; at http://github.com/mmmikael/vlfeat/
Ripal Petel as you mention SIFT requires more computational cost. However, SURF is faster. Both techniques I have implemented but SURF and works well.
There are many such as
Scale-invariant feature transform (SIFT)
Speeded Up Robust Features (SURF)
Binary Robust Invariant Scalable Keypoints (BRISK)
Fast Retina Keypoint (FREAK) ...
Moment Invariants can address your requirements. There is a free Ebook. See the URL http://sciencegatepub.com/books/gcsr/gcsr_vol1/
About combining features,
Wang & Schmid uses a combination of HOG, HOF and MBH in their ICCV2013 paper, "Action Recognition with Improved Trajectories".
A collection of differerent descriptors is compared in "A performance evaluation of local descriptors". A quite famous paper by K.Mikolajczyk and C.Schmid (2005). Here they also introduce the GLOH descriptor
You can use Zernike moments to compute features which are rotationally invariant and can be made easily scale and translation invariant also. Moreover they form very low dimension feature set but little time taking
Dear Ripal Patel,
I am unsure you still need an answer to this question. You may refer to the following paper:
Guohua Lv, Shyh Wei Teng and Guojun Lu, Enhancing SIFT based Image Registration Performance by Building and Selecting Highly Discriminating Descriptors, Pattern Recognition Letters, vol. 84, pp. 156-162, 2016.
Regards,
Guohua