Smoothness-Increasing-Accuracy-Conserving (SIAC) reconstruction, referred to as SR-DG (SIAC Reconstruction-DG), for the discretization of convection-diffusion equations. The method uses high-order SIAC filters over compact multi-element stencils to reconstruct the DG solution and its gradient at quadrature points and element interfaces via a convolution operation. This reconstruction, compared to standard DG, improves inter-element continuity, which enhances the scheme’s performance on coarse meshes, while preserving the global high-order accuracy of the DG discretization. SR-DG method efficiently achieves these objectives and provides additional advantages. It can reconstruct a single state at each interface, reducing the need for numerical fluxes. The method also allows for relaxed CFL constraints on the time step for convection-diffusion problems. Moreover, gradients can be accurately and efficiently obtained from the derivative of the filter, offering an effective alternative to standard DG diffusion schemes. The approach is validated, by means of a systematic comparison with standard DG, through 1D and 2D tests for both linear and nonlinear convection-diffusion problems.