Fig 1.
A taxonomy of object tracking methods.
Fig 2.
The concept of image correlation.
Fig 3.
Correlation pattern recognition.
Table 1.
The proposed video sequences.
Fig 4.
The proposed amoeba video sequences, clear image to the left and annoated image to the right.
a) AMB1 b) AMB2.
Fig 5.
Qualitative assessment of the tested trackers on the a) Tiger1, b) Tiger1_VFM_1, c) Tiger1_VFM_2, d) Surfer, and e) Surfer_VFM sequences [8].
Fig 6.
Qualitative assessment of the tested trackers on the a) Ironman, b) Football1, c) Football1_Modf and d) DragonBaby video sequences [8].
Fig 7.
Qualitative assessment of the tested trackers on the a) Box [8], b) AMB1 and c) AMB2 sequences.
Table 2.
The average fps for the eight trackers.
Fig 8.
Success and precision plots of error and overlap for OPE, SRE, and TRE.
Fig 9.
Attribute-based success plots of SRE (illumination variation, out-of-plane rotation, scale variation, occlusion, deformation, and motion blur).
(a) Illumination variation. (b) Out-of-plane rotation. (c) Scale variation. (d) Occlusion. (e) Deformation. (f) Motion blur. (g) Fast motion. (h) In-plane rotation. (i) Out of view. (j) Background clutter. (j) Background clutter.
Table 3.
The percentage of frames the trackers managed to capture the object from the entire sequence.
Table 4.
The frame ranges of CF-based trackers when failing to capture objects.