High-density objects such as metal prostheses, surgical clips, or dental fillings generate streak-like artifacts in computed tomography images. In this study, we present a novel method for cone-beam CT metal artifact reduction by approximate missing information into the corrupted sinogram. We used sinogram decomposition to compute the structural tensor of the sinogram to estimate the direction. The proposed algorithm consists of two main steps. The first step decomposes the sinogram to estimate rough structure of the sinogram on the metal corrupted part. Using the decomposed sinogram, we estimate the missing sinogram on the second part. We compared the result of our algorithm with metal artifact reduction with different interpolation methods. We showed that our algorithm outperformed other interpolation algorithms such as linear, moving least squares, and adaptive moving least squares algorithms.