MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641098818 A) filed by Manipal Academy Of Higher Education on August 14, 2026, for A System And Method For Visual Feature Compression And Similarity Optimisation Of Digital Images.
Inventors include Bhoomika S V; Deepika C; and Dinesh Rao B.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: A visual feature compression and similarity optimization system and a method, comprising a memory coupled to a processor, the memory storing a feature extraction engine, compressed reference feature vectors, and instructions executable by the processor to: receive and normalize a digital image and a reference image; extract feature vectors using a global average pooling layer; reduce the feature vectors to 50 dimensions using principal component analysis; generate reconstructed images from the reduced feature vectors; evaluate cosine similarity between the reduced feature vectors; perform iterative refinement until convergence or 100 iterations; and output the reconstructed images and 50-dimensional feature vectors corresponding to maximum similarity, thereby achieving 288:1 compression, processing time reduced from approximately 3 minutes to approximately 1 minute, and cosine similarity of 0.9698, a synergistic effect not achievable by any single prior-art method. FIG. 3
Disclaimer: Curated by HT Syndication.
