In a resource-constrained computing environment, it is essential to simplify complex meshes of a huge 3D model for visualization, storing and transmission. Over the past few decades, quadric error metric (QEM) has been the most popular error evaluation method for mesh simplification because of its fast computation time and good quality of approximation. However, quadric based simplification often suffers from its large memory consumption. Since recent 3D scanning systems can acquire both geometry and color data simultaneously, the size of model and memory overhead of quadric increases rapidly due to the additional color attribute. This paper proposes a new error estimation method based on QEM and half-edge collapse for simplifying a triangular mesh model which includes vertex color. Our method calculates geometric error by the original QEM, but reduces the required memory for maintaining color attributes by a new memory-efficient color error evaluation method.