A novel Hilbert transform algorithm for detecting phase from fringe pattern is proposed.Only one interference pattern is required for the phase evaluation.In practice,the Hilbert transform based on the Fourier transformation eliminates all negative frequency components,and any phase variation in the fringe pattern having negative local spatial frequency can not be correctly reconstructed.A procedure for detection of regions having zero spatial frequency in the fringe pattern is proposed by an estimation function for complicated phase distribution.A binary matrix is used for modifying wrapped phase of complicated phase distribution fringe pattern, and a method for removing the bias component by twice Hilbert transform is proposed.The correct values of phase is obtained by an unwrapping process.The performance of the proposed method is evaluated by computer simulation.