In hard decision cooperative spectrum sensing (CSS), each secondary user (SU) or simply cognitive radio user (CR) senses the primary user (PU) activity via a Sensing channel (S-channel) and forwards its own binary decision to a fusion center (FC) via a Reporting channel (R-channel) to make a final decision regarding PU existence. In practical scenarios, both S-channels and R-channels are contaminated with noise, fading and shadowing effects. Thus, the FC may receive faulty decisions from the CRs, which in turn degrades the overall sensing performance of the cognitive radio networks (CRNs). In this paper, an efficient hard decision CSS with two-stage censoring is proposed for boosting the sensing performance of CRNs against noise uncertainty inherent in the S-channels and erroneous inherent in the R-channels. In the first stage, CRs with low quality R-channels are censored by the FC, hence only CRs with high quality R-channels are selected for the next stage of censoring. In the second stage, the low confident CRs with high noise uncertainty factors of their S-channels are censored by the FC, i.e., the FC selects the candidate CRs with the highest quality R-channels and the lowest noisy S-channels. For boosting the sensing decisions made by the CRs, a double dynamic threshold (DDT) is utilized by each CR based on an estimated value of the noise uncertainty factor of its S-channel. The new detection and false alarm probabilities are evaluated mathematically for the proposed scheme. Moreover, numerical analysis is used to confirm the high potency of the proposed scheme over some existing hard decision CSS schemes.