User-Guided Media Object Removal Using Segmentation Masks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for removing distracting objects from visual media items are laborious and often result in incomplete or inaccurate object removal, leading to false positives or incomplete segmentation.

Innovation Solution

A computer-implemented method that translates user input indicating objects to be erased into a bounding box, uses a segmentation machine-learning model to generate a segmentation mask, and optionally employs an inpainting model to remove the objects, ensuring accurate and complete removal based on user intent.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual image or video editing techniques are used to remove distracting objects, then users can remove objects from media items, but the process becomes arduous and time-consuming

Engineering Contradiction:
Improveobject removal accuracyVSAvoidediting time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical editing operations with an automated machine learning system. The segmentation model automatically identifies and segments distracting objects based on user input, eliminating the need for manual pixel-by-pixel editing while maintaining high accuracy through intelligent object recognition and segmentation algorithms.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service object removal by allowing users to simply indicate the general region of distracting objects, and the machine learning model automatically performs the complete segmentation and removal process. This eliminates the need for users to manually edit each object while still providing precise control over what is removed.

Inventive Principle:
Principle #25Self-service

2Productivity

If automatic removal of distracting objects is implemented, then the process becomes faster and easier, but it results in false positives where additional objects are removed or incomplete segmentation

Engineering Contradiction:
Improveobject removal speedVSAvoidsegmentation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system incorporates feedback mechanisms where the segmentation model generates initial segmentation results, which are then refined based on user confirmation or correction. The model learns from user feedback to improve segmentation accuracy, reducing false positives while maintaining high productivity through automated processing of correctly identified objects.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The segmentation process is made dynamic and adaptive, allowing the system to adjust its segmentation thresholds and parameters based on the specific characteristics of each image and user preferences. This dynamic adjustment enables the system to maintain high accuracy across diverse scenarios while preserving fast automated processing capabilities.

Inventive Principle:
Principle #15Dynamics

3Reliability

If a segmentation model processes the entire media item, then complete object detection is achieved, but processing time and computational resources increase significantly

Engineering Contradiction:
Improveobject detection completenessVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies segmentation at multiple levels: first segmenting the image into regions of interest based on user input, then applying the heavy-duty segmentation model only to these relevant regions. This hierarchical segmentation approach ensures complete detection of distracting objects while dramatically reducing processing time by avoiding unnecessary analysis of entire media items.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system applies different processing qualities to different regions of the image. High-accuracy segmentation is applied only to regions containing distracting objects, while other regions receive minimal or no processing. This local quality approach maintains detection completeness for target objects while reducing overall processing time and computational resources.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4309116B1User input based distraction removal in media items
Publication Date: 2026.02.04 GOOGLE LLC
  • EP4309116B1 patent drawingFigure 1
  • EP4309116B1 patent drawingFigure 2
  • EP4309116B1 patent drawingFigure 3A

AI summary

A media application receives user input that indicates one or more objects to be erased from a media item. The media application translates the user input to a bounding box. The media application provides a crop of the media item based on the bounding box to a segmentation machine-learning model. The segmentation machine-learning model outputs a segmentation mask for one or more segmented objects in the crop of the media item and a corresponding segmentation score that indicates a quality of the segmentation mask.