Abstract
Existing methods for removing objects from images, such as manual editing or single-prompt generative artificial intelligence approaches, can be time-consuming or may result in incomplete removal and visual artifacts. A system can provide an automated, iterative process for removing designated object classes from digital images, for example, by employing a heuristic-driven feedback loop that utilizes a generative model. Within each iteration, a probabilistic detection step can ascertain if target objects remain by, for example, querying the model multiple times with varied prompts. If objects are detected, a removal and inpainting operation can be performed. This iterative refinement can enable the system to progressively remove objects and potentially correct errors, which may produce a modified image suitable as a foundation for subsequent creative modification or other generative processes.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Leung, Mira, "Heuristic-Driven Iterative Process for Object Removal in Images Using a Generative Model", Technical Disclosure Commons, (November 19, 2025)
https://www.tdcommons.org/dpubs_series/8899