Structured light three-dimensional (3D) measurement techniques have obtained increasing popularity in the field of industrial automation, inverse engineering, and graphics. Recent literatures show that the imperfect of the imaging system in the structured light 3D metrology will cause the discontinuous-induced measurement artifact (DMA) in the area around the discontinuous edge. Existing DMA reduction methods need to detect the all edges accurately first. This procedure is hard to accomplish when the edges are defocused. Meanwhile, the corrected date in the error area relies heavily on the data in its nearest unaffected area, which makes the corrected data unreliable in some situations. In this work, a flexible deconvolution-based method is proposed to solve the above two problems in this paper. Simulation and experiment show that our proposed method can reduce the Root Mean Square phase/height error of DMA by up to 4 times.