A co-processing scheme of frequency offset and phase noise based on cascade extended Kalman filter (EKF) and block-processed linear Kalman filter (LKF) is proposed. The EKF is responsible for preliminary estimation of frequency offset. The LKF is responsible for tracking frequency offset and phase noise accurately. Relationship between the optimal block length and the tuning parameter Q, linewidth tolerance, frequency offset estimation range and frequency offset tracking speed of algorithm are discussed and analyzed in detail. The results show that the scheme has fast convergence performance, and can achieve high estimation accuracy of frequency offset and phase estimation. Moreover, the frequency offset drift can reach 320 MHz/μs. Compared with traditional blind phase search method, the scheme has high frequency offset tolerance and low implementation complexity. Finally, the carrier recovery performance for quadrature phase shift keying (QPSK) optical communication system is experimentally studied, and the carrier frequency offset estimation performance under different optical signal-to-noise ratios and block data lengths is given.