An urn contains $M$ balls: $W$ white balls, $B$ blue balls and $R$ red balls. A sample $s$ of size $N$ is drawn at random. $M$ is large relative to $N$, and the possibility of picking the same ball twice can be neglected. Number of red balls and number of blue balls in the sample are denoted as $r$ and $b$ respectively. I understand that their variances are as follows:
$$\sigma^2_r=N\frac{R(M-R)}{M^2}\quad \text{and} \quad \sigma^2_b=N\frac{B(M-B)}{M^2}$$
What is covariance between $r$ and $b$, and what is unbiased estimate of the covariance based on sample $s$