MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202641092869 A) filed by Ram Soorat; Gururaj Nase; Veeresh Biradar; P Sripalreddy; Upasana Patil; and Shruti Modi on July 31, 2026, for System And Method For Underwater Image Enhancement Using Channel-Adaptive Frequency-Spatial Deep Learning Network.
Inventors include Ram Soorat; Gururaj Nase; Veeresh Biradar; P Sripalreddy; Upasana Patil; and Shruti Modi.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention discloses a computer-implemented system and method for underwater image enhancement using a coordinated frequency-spatial deep learning architecture. An underwater image acquired through an image acquisition module is preprocessed and converted into shallow feature representations before being separated into low-frequency and high-frequency components using a frequency decomposition module. The low-frequency component undergoes wavelength-aware channel calibration through a channel-adaptive attention block, while the high-frequency component is processed by a spatial feature refinement block employing multi- scale dilated convolutions and spatial attention for structural restoration. The processed feature representations are integrated through a multi-scale feature fusion module and reconstructed by a decoder to generate an enhanced underwater image exhibiting improved color fidelity, contrast, and structural consistency. The disclosed system is deployable on embedded and edge computing platforms and supports real-time processing for autonomous underwater vehicles, remotely operated vehicles, marine inspection, ecological monitoring, and underwater robotic vision applications.
Disclaimer: Curated by HT Syndication.
