A Dataset-free Self-supervised Disentangled Learning Method for Adaptive Infrared and Visible Images Super-resolution Fusion

This study proposes a novel general dataset-free self-supervised learning framework based-on physical model named self-supervised disentangled learning (SDL), and proposes a novel method named Deep Retinex fusion (DRF) which applies SDL framework with generative networks and Retinex theory in infrared and visible images super-resolution fusion. Meanwhile, a generative dual-path fusion network ZipperNet and adaptive fusion loss function Retinex loss are designed for effectively high-quality fusion. The core idea of DRF (based-on SDL) consists of two parts: one is generating components which are disentangled from physical model using generative networks; the other is loss functions which are designed based-on physical relation, and generated components are combined by loss functions in training phase. Furthermore, in order to verify the effectiveness of our proposed DRF, qualitative and quantitative comparisons compared with six state-of-the-art methods are performed on three different infrared and visible datasets. Our code will be open source available soon at https://github.com/GuYuanjie/Deep-Retinex-fusion.

Deep Fusion Prior for Multi-Focus Image Super Resolution Fusion

This paper unifies the multi-focus images fusion (MFIF) and blind super resolution (SR) problems as the multi-focus image super resolution fusion (MFISRF) task, and proposes a novel unified dataset-free unsupervised framework named deep fusion prior (DFP) to address such MFISRF task. DFP consists of SKIPnet network, DoubleReblur focus measurement tactic, decision embedding module and loss functions. In particular, DFP can obtain MFISRF only from two low-resolution inputs without any extent dataset; SKIPnet implementing unsupervised learning via deep image prior is an end-to-end generated network acting as the engine of DFP; DoubleReblur is used to determine the primary decision map without learning but based on estimated PSF and Gaussian kernels convolution; decision embedding module optimizes the decision map via learning; and DFP losses composed of content loss, joint gradient loss and gradient limit loss can obtain high-quality MFISRF results robustly. Experiments have proved that our proposed DFP approaches and even outperforms those state-of-art MFIF and SR method combinations. Additionally, DFP is a general framework, thus its networks and focus measurement tactics can be continuously updated to further improve the MFISRF performance. DFP codes are open source and will be available soon at http://github.com/GuYuanjie/DeepFusionPrior.