Hardware feasibility analysis for motion segmentation initialization
Research output: Contribution to journal › Conference article › Research › peer-review
We analyze and evaluate different initialization methods for motion layers segmentation, a powerful mid-level vision tool. We estimate 6-D affine motion models from the optical flow corresponding to different over-segmentations. Over-segmentations can be either uniform rectangular blocks; or adaptive sized rectangular blocks; or super-pixels; or based on any other clustering methods. We present performance analysis of motion segmentation initialization algorithms on video sequences and discuss the relative pros and cons of these methods in terms of hardware implementation issues.
Original language | English |
---|---|
Journal | ISOCC 2014 - International SoC Design Conference |
Pages (from-to) | 148-149 |
Number of pages | 2 |
DOIs | |
Publication status | Published - 16 Apr 2015 |
Externally published | Yes |
Event | 11th International SoC Design Conference, ISOCC 2014 - Jeju, Korea, Republic of Duration: 3 Nov 2014 → 6 Nov 2014 |
Conference
Conference | 11th International SoC Design Conference, ISOCC 2014 |
---|---|
Country | Korea, Republic of |
City | Jeju |
Period | 03/11/2014 → 06/11/2014 |
Bibliographical note
Publisher Copyright:
© 2014 IEEE.
- affine motion, layered representation, optical flow, superpixel, video segmentation
Research areas
ID: 301829472