Image motion estimation


Image a	Image b	Image c	Image d

Original Image	Moved Image	Real Rotation degrees	Real Translation (x,y) pixels	Estimated Rotation degrees	Estimated Translation (x,y) pixels	Error Rotation degrees	Error distance pixels
a	b	-11.31	(33.30,-3.03)	-11.25	(34.96,-2.97)	0.06	1.66
a	d	-12.53	(9.93,-3.37)	-11.95	(11.47,-2.81)	0.58	1.64
b	c	0.00	(19.00,-3.00)	0.00	(19.00,-3.00)	0.00	0.00

The following sequence shows the selection and the stabilization of a pertinent image region.
The fist column, images from (a) to (e), presents the sequence of perceived images. Figure (f) shows the automatic selection of pertinent region, based on a scene interpretation task and, figure (k) shows the corresponding selected sub-image from the image (a).
The second column, from image (g) to (j), shows the sub-images extracted from perceived images. These images show the perturbations induced from the sensor motion.
The stabilization performed over the sub-images is shows in the third column. These sub-images always show the same region over the original images. The motion between a couple of images is used to correct the perturbation in the next image of the sequence.

Extracted Original Image	Extracted Moved Image	Real Rotation in degrees	Real Translation (x,y)in pixels	Estimated Rotation in degrees	Estimated Translation (x,y) in pixels	Error Rotation in degrees	Error distance in pixels
k	g	-0.50	(-0.71,28.60)	0.00	(0.00,28.00)	-0,50	0.92
l	h	-0.00	(7.00,2.00)	0.00	(7.00,2.00)	0.00	0.00
m	i	0.00	(3.00,30.00)	0.00	(3.00,30.00)	0.00	0.00
n	j	0.00	(13.00,22.00)	00.00	(12.00,22.00)	0.00	1.00

Motion quantification