Suitability of a particular tracking method largely depends on image/video condition. If segmentation of targets is easier then the simplest method would be applying nearest neighbor to link/associate those segmented objects in subsequent frames. You can also combine temporal tracking information to correct segmentation results. Active contour based tracking is an alternative simple way, that can be adapted to both segmentation based or model based tracking where model based tracking tracking performs segmentation and tracking at the same time. 

Dear Aashay,

do you need object tracking algorithms for image applications ?

In this case, do you track objects in the measurement space (i.e. your image / video sequence) or in a state space (e.g. position in a reference frame).

You can try using Template Matching.

I have done a study on mean shift tracking in my graduation thesis. It's not so difficult. If you want, i can show it to you.

The easiest to implement is probably template tracking using normalized cross-correlation. Success depends on the saliency of your object and across how many frames it needs to be tracked.  See normxcorr2 if you're using MATLAB.

Hello. My understanding is that the mean shift algorithm tries to fit a Gaussian density to the image and get the mode. I'm not even sure if this is tracking since the usual definition assumes that you have prior information (usually from the previous iteration of the algorithm) on the state PDF (mean, covariance, etc) of the object. Depending on your problem, you may be able to use something simple like thresholding to get the centroid of the object to be tracked. You can then use any one of a number of established tracking techniques, the simplest of which is the alpha-beta nearest neighbour filter. For a review of such approaches, please see the following link.

Article Taxonomy of multiple target tracking methods

Hello Aashay,

We have recently published a paper that contributes a new algorithm for people tracking. It is very simple and it is based on easy concepts such as LSAP and Kalman filtering. Furthermore, we have obtained remarkable results in human tracking with a good efficiency, as you can see:

http://dx.doi.org/10.1016/j.eswa.2015.06.016

https://www.youtube.com/watch?v=Rqk-vFKzAcQ

I hope that this contribution can help you. Good luck with your research!

Article Expert Video-Surveillance System for Real-Time Detection of ...