Ultrasonic phased array total focusing method (TFM) has been widely used in the field of non-destructive testing (NDT) as an advances post-processing imaging algorithm. It shows great potential in defect detection and characterization. However, due to the small number of excitation elements and large amount of collecting data, its temporal resolution and imaging efficiency need to be improved. To overcome these disadvantages, an ultrasonic phased array total focusing method based on sparse deconvolution has been proposed in this paper. Based on the actual situation of ultrasonic testing, a regularized sparse deconvolution model has been established. The Bayesian theory is used to add low-probability prior information to the deconvolution model. At the same time, the sparse deconvolution optimization algorithm proposed in this paper is used to solve the sparse deconvolution model. Compared with ordinary phased array total focusing method, this algorithm reduced the calculation amount and improved the temporal resolution. The performance of the proposed method is proved by simulation and experimental comparison. The results verify that the proposed method has higher resolution and imaging efficiency.