Noise power estimation is a key issue in modern wireless communication systems. It allows resource allocation by detecting white spectral spaces effectively, and gives control over the communication process by adjusting transmission power. Thus far, the proposed estimation methods in the literature are based on spectral averaging, eigenvalues of sample covariance matrix, information theory, and statistical signal analysis. Each method is characterized by certain stability, accuracy and complexity. However, the existing literature does not provide an appropriate comparison. In this paper, we evaluate the performance of the existing estimation techniques intensively in terms of stability and accuracy, followed by detailed complexity analysis. The basis for comparison is signal-to-noise ratio (SNR) estimation in simulated industrial, scientific and medical (ISM) band transmission. The source of used background distortions is complex noise measurement, recorded by USRP-2932 in an industrial production area. Based on the examined solutions, we also analyze the influence of noise samples separation techniques on the estimation process. As a response to the defects in the used methods, we propose a novel noise samples separation algorithm based on the adaptation of rank order filtering (ROF). In addition to simple implementation, the proposed method has a very good 0.5 dB root-mean-squared error (RMSE) and smaller than 0.1 dB resolution, thus achieving a performance that is comparable with the methods exploiting information theory concepts.