The research of visible surface computation has been being done for a long time. The purpose of this research is to produce a complete description of a surface that is only partially constrained by available data. The problem to find a visible-surface depending on the data given has been formulated in a regularization frame-work by D. Terzopoulos and others. The result of the desired surface is a solution of a partial differential equation. When discretized with finite element to technique, this equation gives rise to a large sparse linear system. To solve this large sparse linear system takes very long time. In their method, the rectangle is used as the basic element to discretize the working area with finite elment technique. In fact, in many cases there are many intersection points in the mesh where are no available data. This is one of the reasons that make the sparse linear system very large. And in many cases, the available data are distributed randomly to a degree in the working area. In those cases, it is difficult to find a uniform rectangular mesh that can make all of the available data on its intersection points or make the mesh in a very small scale when the available data are in a quite number. In order to solve this problem, here we propose to use triangle as the basic element to replace the rectangle to discretize the working area. We will only use the points that the available data exist as the vertices of the triangles. So we can reduce the size of the sparse linear system to save processing time. The remainder of this paper is organized as the following : Section 2 discusses the shape function on a triangular element. Section 3 expresses the implementation of the computation we propose, and Section 5 concludes with discussions about future work.