It is a very basic question, but I am confused in it.
As my thesis title is 'Development of Video segmentation techniques
It is well known that, there are two types of video segmentation techniques (1) spatial segmentation (2) temporal segmentation, and third, we can also say spatio-temporal segmentation.
Now my first question is that:
If we are performing motion based video object segmentation i.e. temporal segmentation (or we can say we are detecting or tracking the moving objects), then will it video segmentation?
Now second question is: (continue with first question)
or we have to perform the spatial segmentation also to completely do the video segmentation (i.e. is it mandatory to perform the spatio-temporal segmentation?)
This question may look silly. However it is important for me. Expert in this field kindly reply.
This is a very good question.
A good way to begin answering this question is to consider the following paper:
A Survey on Video Segmentation
https://www.researchgate.net/publication/283532444_A_Survey_on_Video_Segmentation
Hampapur, A., Jain, R., Weymout, T.E. observe that video segmentation is performed based on the types of editing operation that an editor can perform when editing two shots (p. 910).
See, also:
Point cloud video object segmentation using a persistent supervoxel world-model
https://www.researchgate.net/publication/261353299_Point_cloud_video_object_segmentation_using_a_persistent_supervoxel_world-model
and
One-Shot Video Object Segmentation
https://www.researchgate.net/publication/310440999_One-Shot_Video_Object_Segmentation
Conference Paper Point cloud video object segmentation using a persistent sup...
Conference Paper A Survey on Video Segmentation
Article One-Shot Video Object Segmentation
Not an expert Sandeep , now as you have mentioned segmentation broadly can be done on spatial data (correlation) and temporal . The idea is to achieve greater compression and reduce reconstruction efforts (because of lossy) .
Now , if you consider motion tracking the static field is say background. If the subject moves from this scene to another with many objects that move then temporal segmentation (or boundaries change) so does the spatial .
Eg: H.264 (why there are finite set of P I B ) frames .
So yes , it is to improve greater efficiency for packing the data vs decoding efforts (with least amount delay while rendering ) since it is live ...
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing