We propose an image resolution improvement method for optical coherence tomography (OCT) based on sparse continuous deconvolution. Traditional deconvolution techniques such as Lucy-Richardson deconvolution suffers from the artifact convergence problem after a small number of iterations, which brings limitation to practical applications. In this work, we take advantage of the prior knowledge about the sample sparsity and continuity to constrain the deconvolution iteration. Sparsity is used to achieve the resolution improvement through the resolution preserving regularization term. And the continuity based on the correlation of the grayscale values in different directions is introduced to mitigate excessive image sparsity and noise reduction through the continuity regularization term. The Bregman splitting technique is then used to solve the resulting optimization problem. Both the numerical simulation study and experimental study on phantoms and biological samples show that our method can suppress artefacts of traditional deconvolution techniques effectively. Meanwhile, clear resolution improvement is demonstrated. It achieved nearly twofold resolution improvement for phantom beads image that can be quantitatively evaluated