Abstract: Under the framework of L1-norm based Super-Resolution Reconstruction(SRR), by introducing a method for automatically estimating regularization parameter, a SRR method based on mixture prior model is proposed according to difference image statistics combined with probability distribution model. Experimental results indicate the proposed approach which results in good robustness and effective convergence outperforms the old one.

Key words: super-resolution, difference image, mixture prior model

摘要： 在L1范数图像超分辨率重建算法框架下，引入参数自适应估计，结合差分图像统计特性和概率分布模型提出一种基于混合先验模型的超分辨率重建方法。实验证明该方法可以弥补L1范数重建方法的不足，获得更多的图像细节，对模型误差表现出良好的稳健性，可以加速收敛。

关键词: 超分辨率, 差分图像, 混合先验模型

CLC Number: 

• TP393