This paper considers the issue of noise reduction associated with Optical Feedback Self-mixing Interferometries (OFSMI). The objective is to develop an effective pre-processing filtering solution that can eliminated the inherent noise disturbances. The characteristics of OFSMI signals are described in some detail and two filter solutions are proposed. The latter includes a non-linear median filter and a Kaiser based FIR filter. The performance of the two types of filters are investigated. It is shown that median filters are capable of removing sparkle-like noise while Kaiser based FIR filters are effective in reducing the high frequency noise as well as the slow time-varying signal envelop fluctuation. Interestingly our results show that the best performance is achieved by combining the two filters, that is, a median filter followed by a Kaiser-based FIR filter. In other words, the estimation accuracy of OFSMI parameters such as the line-width enhancement factor (LEF) of semiconductor lasers is significantly improved with the aid of the proposed pre-processing solution. ©2007 IEEE.