In this study, a text interference elimination algorithm is proposed based on the total variation (TV) model with respect to the visual influence of the text in an image. Roberts operator and binary morphology are applied to detect and accurately locate the interferential text area in an image. Subsequently, the interferential text can be effectively eliminated by considering the determined interferential text area as the constraint condition of the TV model and solving it. The simulation experiments denote that the proposed method is efficient and practical for the detection and elimination of interferential text from images.