MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202621048566 A) filed by Shrinivas Sachin Patil; Swapnil Harishchandra Khambe; Nrusinha Vitthal Mane; Sanskar Rajesh Patil; and Prof. Vijaykumar Raghunath on April 16, 2026, for Hybridrecon: Adaptive Hybrid 3d Reconstruction System Using Neural And Multi-View Techniques With Automated Quality Assessment.
Inventors include Shrinivas Sachin Patil; Swapnil Harishchandra Khambe; Nrusinha Vitthal Mane; Sanskar Rajesh Patil; and Prof. Vijaykumar Raghunath Ghule.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: This study proposes an intelligent agent-based system that aims to classify multi-view images for 3D reconstruction automatically. The procedure includes the following steps: First, the perceptual component examines each image, scoring it on sharpness, features density, texture, luminosity, and contrast. Next, the complete set of images is evaluated based on several criteria such as the number of perspectives and subject homogeneity. Lastly, the gating component combines all scores to identify which actions must be taken, either PROCEED, WARN, or REJECT Prior to the execution of further steps, the image undergoes preprocessing by passing through a thirteen-stage pipeline. The process comprises duplicate identification using perceptual hashing, resolution-based blur removal, noise reduction, histogram equalization, and foreground extraction. Once the mesh is created through the 3D reconstruction process, the confidence model tests whether the mesh obtained from the process is good enough. Thereafter, the prediction made by the system concerning suitability for reconstruction is evaluated against the output of the process using the Pearson correlation coefficient.
Disclaimer: Curated by HT Syndication.
