(1a) Circle Image Sequence (with blurred/edge detected version on the right)
(1b) Circle Image Sequence with Noise (with blurred/edge detected version on the right)
(2) House Image Sequence
(3) Heart Image Sequence
We tested our program with a set of test images that were increasing in difficulty of region detection.
We first tested our snake algorithm with single static images of a rectangle and an ellipse. These were chosen particularly because they approximated our tracked objects in our image sequences and were also an opportunity for us to fine tune the parameters to be used for our various region detection methods, and especially to determine to values for α, β, and γ. Once these test images verified that our snake algorithm was working well, we proceeded to image sequences.
We first tracked a simple image sequence of a white circle slowly becoming a thin dark gray ellipse. We also tried testing on the same image but with added noise. Then, we tested our program with determined parameter values on two real-life image sequences. Sequence 1 involves tracking the roof of a rotating house, where we wanted to test the efficacy of the snake in tracking an angular object through a relatively drastic shape transformation with quite significant shift in distance. Sequence 2 involves tracking the heart as it beats, where this time we wanted to test the efficacy of the snake in handling a relatively odd shaped object and significant noise.