MUMBAI, India, Feb. 13 -- Intellectual Property India has published a patent application (202641008780 A) filed by Mrs. B. Sankaralakshmi; Mrs. S. Pradeepha; Dr. S. V Anandhi; and Dr. R M Rajeshwari, Rajapalayam, Tamil Nadu, on Jan. 28, for 'hybrid image super resolution using wavelet and transformer networks.'
Inventor(s) include Mrs. B. Sankaralakshmi; Mrs. S. Pradeepha; Dr. S. V Anandhi; and Dr. R M Rajeshwari.
The application for the patent was published on Feb. 13, under issue no. 07/2026.
According to the abstract released by the Intellectual Property India: "Applications such as medical imaging, remote sensing, surveillance, and multimedia processing depend on image super-resolution, which seeks to reconstruct high-resolution images from low-resolution sources. The limitations of conventional interpolation methods and convolutional neural network approaches in capturing global contextual information and intricate structural details often result in artifacts and indistinct edges. Despite recent successes in capturing long-range correlations, Transformer-based models are computationally intensive and may inadequately preserve information at high frequencies. This research presents a hybrid framework for picture super-resolution that integrates Transformer neural networks with wavelet-based multi-resolution analysis. The proposed method utilizes a discrete wavelet transform (DWT) to efficiently isolate structural and detail components by initially reducing the low-resolution input image into multiple frequency sub-bands. A Transformer-based self-attention network is employed to enhance the decomposed wavelet features. This network selectively enhances high-frequency information while preserving global contextual relationships. The enhanced features are utilized to reconstruct a high-resolution image by the application of an inverse discrete wavelet transform (IDWT). The experimental findings indicate that the proposed hybrid framework surpasses conventional CNN- and Transformer-exclusive super-resolution methods for edge preservation, texture reconstruction, and artifact mitigation. The proposed method is highly appropriate for numerous real image enhancement tasks, owing to its favorable balance between computational efficiency and reconstruction precision. Keywords: Image Super-Resolution, Wavelet Transform, Discrete Wavelet Transform (DWT), Transformer Networks, Self-Attention Mechanism, Hybrid Deep Learning, Frequency-Domain Feature Learning)."
Disclaimer: Curated by HT Syndication.
