Neural Face Image Processing for Covering Object Removal

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing face recognition systems are affected by covering objects such as glasses, makeups, scars, and tattoos, which can impair the accuracy of matching scores and recognition results.

Innovation Solution

A neural network-based image processing method that utilizes a multi-layered structure of sub-neural networks to process images, allowing for the removal of covering objects by generating an image of the face without these obstructions based on the image's features.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If face recognition is performed directly on images with covering objects, then the recognition process is simple and fast, but the matching score accuracy deteriorates

Engineering Contradiction:
Improverecognition process speedVSAvoidmatching score accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by performing covering object detection and removal before the face recognition process. The system identifies covering objects in the input image, generates a mask to cover these objects, and then performs recognition on the masked image. This preliminary processing step ensures that the recognition process operates on cleaned images, improving matching score accuracy without significantly impacting overall system efficiency.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If covering objects are removed before recognition, then the matching score accuracy is improved, but the device complexity increases

Engineering Contradiction:
Improvematching score accuracyVSAvoidimage processing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the image processing task into distinct modules: a covering object detection module that identifies covering objects, a mask generation module that creates coverage masks, and a face recognition module that performs identification. This modular segmentation allows each component to be optimized independently and facilitates easier system maintenance and deployment while achieving accurate covering object removal.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a mask as an intermediary element between the input image and the face recognition process. The mask selectively covers detecting features of covering objects while preserving the underlying face information, acting as a mediator that enables accurate recognition without requiring complex direct removal of covering objects. This intermediary approach simplifies the overall system architecture.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Loss of time

If covering objects are removed by traditional methods, then the processing time is short, but the realism of generated covered face parts deteriorates

Engineering Contradiction:
Improveprocessing timeVSAvoidrealism of generated covered face parts
Core Design Contradiction:
Loss of timeVSManufacturing precision

Solution Approach 1:

The patent applies copying by generating a mask that replicates the covering regions and using this mask to guide the recognition process. Instead of attempting to reconstruct or remove covering objects through complex image processing, the system creates a simplified copy (mask) that represents the covering areas, which is then used to adjust the recognition output. This copying approach maintains processing efficiency while improving the realism of the final recognition result.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3698269B1An image processing method and system
Publication Date: 2025.07.30 ZHEJIANG DAHUA TECH CO LTD
  • EP3698269B1 patent drawingFigure 1
  • EP3698269B1 patent drawingFigure 2
  • EP3698269B1 patent drawingFigure 3

AI summary

A neural network-based image processing method may include receiving, by a trained neural network, a first image including a first object, the first object being partially covered by a second object. The method may also include generating, by the trained neural network, a second image based on the first image. The second image is a representation of the first image with the second object substantially removed, and the first object is a human face.