Infrared Image Denoising via L1/2 Sparse Representation over Learned Dictionary

Infrared (IR) images often have low resolution and vague details, resulting in lower image quality and poor visual effect. This paper comes up with an Infrared image denoising method via L_1/2 sparse representation, while simultaneously training a over-complete dictionary on its content using the K-SVD algorithm. Experiment results have shown excellent denoising ability of the proposed denoising method, which can efficiently reduce Gaussian noise while exploiting much more image texture information.