Computer Sciences and Information Technology
Title : | Investigation of Clustered Anisotropic Diffusion Model and Multi-structured Weighted Networks for Fusion of Multimodal Images |
Area of research : | Computer Sciences and Information Technology |
Focus area : | Image Processing and Computational Mathematics |
Principal Investigator : | Dr. Arpita Das, University Of Calcutta, West Bengal |
Timeline Start Year : | 2024 |
Timeline End Year : | 2027 |
Contact info : | adrpe@caluniv.ac.in |
Details
Executive Summary : | Present work proposes a new concept of multimodal image fusion technique using some supervised/unsupervised clustering models with appropriate feature selection algorithms. The clustering approach helps us to separate out various uncorrelated information contents of the source images into different clusters. To reduce high frequency edge distortion and chromatic degradation of the fused data, anisotropic diffusion filter may be used as a feature selection algorithm. This filtering technique can produce a coarse approximation of the source data which in turn helps us to refine gross-textural (GT) as well as gross-edge (GE) information of the source images. Hence the fine-textural (FT) and fine-edge (FE) information of different clusters can be further filtered out by subtracting GT and GE from the source data respectively. Following this decomposition and feature selection algorithm, some weighted network model based fusion rule may be introduced to integrate different gross- and fine-structured components of the source images. The weight factors of this network model can be estimated in terms of the local image statistics and hence no pre-training data is required for learning purposes. In this view, the proposed fusion rule is declared to be faster than any deep learning model based fusion approach. The importance of the proposed fusion methodology can be summarized as: (a) It applies some fundamental mathematical understandings to meet the challenges of fusion technique, (b) It will produce good fusion outcomes irrespective of the nature of the multimodal images. |
Total Budget (INR): | 6,60,000 |
Organizations involved