Kalman filter Vs averaging Assume I have 2 sensors that measure the same output but with different SNR
which approach will give me better sensing?
1.  avaearging the two sensores?
2.  use Kalman filter on one of them ( assuming on the worse SNR)
 A: It depends on your problem. When using averaging you can not propagate the variances. I would prefer a simple Kalman filter involving both sensor measurements.
You just need to proceed the update step twice:
init()
predict()
update(sensor1)
update(sensor2)
predict()
update(sensor1)
update(sensor2)
and so on...

So you can have an estimate of your current variance and (if you want) smooth the sensor readings by means of the Q matrix.
Have a look at the following post. The problem was pretty the same. One needed to fuse 4 position signals.
multi-sensors fusion using Kalman filter
UPDATE
About variance propagation: 
If you just use an averaging you weigh the measurements based on their variances (which can be calculated from the SNR). The measurement with a bigger variance has less impact on the resulted value. But you can not involve the variances from the earlier measurements. You just have a new variance to the current point in time. The averaged value at the time t-1 has also no impact on the value at the time t.
If you have two measurements $X_1 \sim N(\mu_1, \sigma_1)$ and $X_2 \sim N(\mu_2, \sigma_2)$ then the fused value would be:

When using a kalman filter you can estimate the current state $x_k$ and its covariance $P_k$ with respect to the previous state $x_{k-1}$ and its covariance $P_{k-1}$ involving the new measurement $y$ and its variance $R$. The covariance can also be increased during the prediction step using the system noise information $Q$.
Here is the math for means and covariances on different steps of the kalman filter (assuming there is no state transition $F = I$):
First prediction after the initialization

Update with the first measurement $meas1$

Update with the second measurement $meas2$

Prediction

and so on...
The variance is being propagated. It goes from state to state and gives you a better information about the state uncertainty at any point in time.
So how you can see the kalman filter lets you combine all the information about the system and the measurement. The measurements are not being processed as isolated numbers any more.
