Rakshana96/Fusion-of-Infrared-and-Visible-Images-for-Surveillance-Applications

A novel image fusion technique is presented for integrating infrared and visual images. Integration of images from the identical or various sensing modalities can deliver the specified information that can't be delivered by viewing the sensor outputs individually and consecutively. This work proposes an Artificial Neural Network (ANN) based imag...

A novel image fusion technique is presented for integrating infrared and visual images. Integration of images from the identical or various sensing modalities can deliver the specified information that can't be delivered by viewing the sensor outputs individually and consecutively. This work proposes an Artificial Neural Network (ANN) based image fusion technique using Gaussian smoothness and Joint Bilateral Filter. The Gaussian smoothness and the Joint Bilateral Filter decompose the source images. The implementation involves the elimination of fine-scale information with Gaussian filtering, extraction of edge and structure with joint bilateral filtering iteration. The decomposition has edge-preserving and scale-aware properties to enhance the acquisition of detail layer. Two layers of rules are used to combine the decomposed layers and the combined layers are used to reconstruct the final fused image. An enhanced fused image is obtained through ANN for a better visual understanding and target detection. The proposed framework is evaluated using quantitative metrics such as Standard Deviation, Edge Dependent Fusion Quality Index, Entropy, Mean Square Error, Peak Signal to Noise ratio, Naturalness Image Quality Evaluator and Structural Similarity Index. This work outperforms visually as well as quantitatively and achieves better performance with the reduced complexity.