Multistep Interactive Image Generation Utilizing Knowledge Store

Ghulam Ahmed Ansari, Sai Koti Reddy Danda, Srikanth G. Tamilselvam, Amrita Saha

Abstract: This patent introduces an AI-powered interactive image editing system that leverages a knowledge store for context-aware image manipulations. Unlike traditional image editing tools with fixed operations, this system analyzes the context of an image, identifies existing elements and their attributes, and uses a machine-learning-driven knowledge store to suggest appropriate modifications. Users can add, move, resize, or recolor objects, with the system intelligently adjusting attributes (e.g., relative size, color) based on real-world constraints. The system iteratively refines edits through user feedback, reducing manual adjustments and making image generation more intuitive and efficient.