In this paper, we study the problem of dynamic spectrum sensing and access under the scenario of correlated subband occupancy and POMDP model in cognitive radio networks. The goal is to sense and access the orthogonal frequency division multiplexing (OFDM) based channels with correlations and the channels' occupancy states are modeled as POMDP. We proposed a novel spectrum sensing and access mechanism based on the joint relevance of time domain and frequency domain. A maximum a posteriori (MAP) estimator of channel occupancy based on measurements from multiple subbands is applied to frequency domain sensing algorithm. For time domain, we adopt the partially observable markov decision process (POMDP) model and present a novel spectrum selection mechanism. Simulation results shows that the proposed spectrum sensing and access mechanism can significantly enhance the sensing performance and increase the throughput of the cognitive radio networks.