The Monte Carlo method is a stochastic statistic algorithm. It is one of the most accurate dose calculation methods, but its application in clinic is limited because of the long computation time. Generally, to accelerate Monte Carlo simulation and reduce stochastic noise, a digital filtering technique is used to smooth a rough dose distribution (includes evident noise) to a satisfied one. Different types of filters have been applied, such as Gaussian filters, Savitzky-Golay filters, etc., but the ability of a single filtering filter is limited. Therefore, a hybrid filter combining those filtering techniques was used. Two types of mixture methods - parallel and cascade - with three-dimensional Gaussian and Savitzky-Golay filters were researched. In addition, a method that simplifies the mixture filter structure using an equivalent convolution kernel based on convolution theory was introduced. With simulation data from a standard phantom, the rough dose distributions and the dose distribution smoothed by the two types of mixture filters were compared with that of the "benchmark" one. Test results showed that the two types of mixture filters can suppress much of the noise added in Monte Carlo dose distributions and enhance its visualization. As for the research's test cases, the filtering effect of the cascade mixture filter was slightly better than that of the parallel mixture filter. Filter combinations can provide favorable filtering effects. The filtering effects of different mixture methods are not uniform. The cascade mixture filter has a better filtering effect than the parallel mixture filter.