MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641068890 A) filed by S. Nandini; and M. Nagabushanam on June 01, 2026, for Adaptive Multimodal Image Segmentation System Using Cross-Modal Deep Feature Fusion.
Inventors include S. Nandini; and M. Nagabushanam.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: [045] The present invention discloses an adaptive multimodal image segmentation system using cross-modal deep feature fusion. The system receives two or more image modalities corresponding to a common target region and processes the images through preprocessing, harmonisation, and cross-modal registration. A plurality of modality-specific deep feature encoders extract hierarchical features from each image modality, while an adaptive cross-modal fusion engine determines region-wise modality reliability and combines the extracted features according to contextual relevance. A segmentation decoder generates a pixel-wise, voxel-wise, or region-wise segmentation output, and a boundary refinement module improves contour accuracy of the segmented region. The system may further generate confidence maps, uncertainty maps, registration reliability values, and modality contribution maps for technical verification of the segmentation output. The invention improves segmentation accuracy under conditions of modality mismatch, noise, misalignment, missing modalities, and variable imaging quality, thereby providing a robust technical solution for medical imaging, remote sensing, industrial inspection, robotic vision, and other multimodal image analysis applications. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
