Table 1 Meanshift algorithm for finding the maximization of probability distribution

Step 1. Compute the distribution p _u in the current frame with y₀ and compute the Bhattacharyya distance (= ρ) between p _u and q _u as follows:

$\rho \left(p\left(y_0\right),q\right)=\sum _{u=1}^n\sqrt{p_u\left(y_0\right)q_u}.$

Step 2. Compute the weight (w _i ) is defined as follow:

$w_i=\sum _{u=1}^m\delta \left[b\left(x_i\right)-u\right]\sqrt{\frac{q_u}{p_u\left(y_0\right)}},$

where b(x _i ) is the color bin of the color at x _i .

Step 3. Derive the new location y₁ can be defined as follow:

$y_1=\frac{\sum _{i=1}^n{x}_i{w}_{i}g\left({∥\frac{y_0-x_i}{h}∥}^2\right)}{\sum _{i=1}^n{w}_{i}g\left({∥\frac{y_0-x_i}{h}∥}^2\right)},$

where h is a window radius and g() is a kernel G.

Step 4. Compute the Bhattacharyya distance (= ρ) between p _u and q _u in the new location y₁.

Step 5. While ρ|p(y₀), q| <ρ|p(y₁), q|

Do y₁ = 1/2(y₀+y₁).

Step 6. If ||y₁-y₀|| <ε Stop.

Otherwise, y₀ = y₁ and go to Step 1.