针对匹配追踪算法在过完备字典库中稀疏表示计算量大的问题,提出一种将遗传匹配追踪算法(GMP)与图像修复相结合以实现修复的方法.该算法采用GMP寻找最佳原子,改进选择算子,将精英保留策略、锦标赛选择方法及轮盘赌方法相结合.将混合选择算子应用到基于GMP的图像修复算法中,通过实验结果表明:该算法具有一定的可行性,并且修复效果很好.
For matching pursuit algorithm sparse repersentation in over-complete dictionary computationally intensive problem, an inpainting algo- rithm combined with genetic matching pursuit(GMP) is proposesd. In order to find the best atom and to improve selection operator,the algorithm a- dopts GMP, and then combines with the elitist strategy, tournament selection and roulette method. The mixed selection operator is applied to inpainting algorithm based on GMP. As a consequence, the experiment shows that the proposed algorithm is feasible in some extent and the effect of inpainting is quite good.