The performance of cooperative spectrum sensing (CSS) with double threshold and censoring of cognitive radios (CRs) has been evaluated using maximal ratio combining (MRC) in the presence of Rayleigh faded sensing (S) and reporting (R) channels. Each CR senses the primary user through sensing channel and evaluates the energy of the received signal at its sensing channel input and performs double threshold based spectrum sensing. The qualified CRs are censored and only the selected CRs report their decisions to fusion centre. Censoring of CRs is done on the basis of the quality of R-channel. Typically two types of censoring such as rank-based and threshold-based censoring are considered to select the best CR based on the quality of the R-channels. The censoring of CRs not only improves the detection performance but also reduces the transmission overhead. The performance of CSS has been investigated in terms of average missed detection probability, total error probability, average number of transmitted bits and agility gain of the network with respect to relevant network parameters such as number of selected CRs, censoring thresholds, fail sensing probabilities and S and R-channel signal to noise ratios under both the perfect and imperfect channel estimation. The performance has been studied for hard decision fusion rule (majority) and soft decision fusion rule (MRC). The impact of channel variation rate on the sensing performance is also indicated. Further the performance of double threshold based CSS with censoring is compared with that of single threshold based CSS with censoring. © 2015, Springer Science+Business Media New York.