Image Processing Apparatus Candidate Selection for Facial Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Users face difficulties in recognizing objects within images, especially in crowded scenes or with poor capturing conditions, such as small or non-frontal facial images, making it challenging to accurately identify individuals using existing facial recognition technologies in consumer-grade cameras.

Innovation Solution

An image processing apparatus that detects selection candidate images, such as facial images, and overlays them as a candidate presentation list on the main image, allowing users to easily select and emphasize the target object for focus, using a CPU or DSP for processing and a display unit for visualization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If individual identification technology is applied to track specific persons, then the ability to identify individuals is improved, but the processing ability requirement becomes excessively high for consumer appliances

Engineering Contradiction:
Improveindividual identification accuracyVSAvoidprocessing ability requirement
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the complex individual identification task into simpler components: detecting facial images, extracting candidate regions, and displaying them in a list. This breakdown reduces the processing burden on consumer-grade devices while maintaining identification capability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary display list that shows detected facial images and their positions. This intermediary representation allows users to identify targets without requiring the system to perform complex real-time individual identification, thus reducing processing requirements.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If facial recognition is performed under various capturing conditions, then the适用范围 is improved, but the identification accuracy deteriorates due to small images, shielding, shadows, and non-frontal faces

Engineering Contradiction:
Improvecapturing condition adaptabilityVSAvoididentification accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent performs partial facial recognition by detecting only sufficient features to identify candidate regions and display them in a list. It does not attempt full individual identification under all conditions, thereby maintaining acceptable accuracy while adapting to various capturing conditions including small images and non-frontal faces.

Inventive Principle:
Principle #16Partial or excessive action

3Area of stationary object

If the entire image is displayed without assistance, then the complete scene is visible, but the user cannot quickly recognize specific objects to focus on

Engineering Contradiction:
Improvedisplay areaVSAvoidobject recognition ease
Core Design Contradiction:
Area of stationary objectVSEase of operation

Solution Approach 1:

The patent segments the display into two parts: the main image showing the complete scene and a separate candidate list showing detected facial images with position information. This segmentation allows users to see both the overall scene and specific targets of interest simultaneously.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent creates a copy of the detected facial images and their position information in a separate display list. This copy provides additional information about where to look in the main image without obscuring the original scene, making it easier for users to locate specific objects.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP2962455B1Image processing apparatus, image processing method, and program
Publication Date: 2022.04.27 SONY GROUP CORP
  • EP2962455B1 patent drawingFigure 1
  • EP2962455B1 patent drawingFigure 2A~2B
  • EP2962455B1 patent drawingFigure 3

AI summary

An information processing apparatus may include a control device. The control device may generate display data including a main image for display in a first area of a display area and a list of candidate images for display in a second area of the display area at a same time the main image is displayed in the first area, wherein the candidate images are extracted from and represent respective objects in the main image. Also, the control device may receive an input from a user to select one candidate image from the list, wherein the display data is generated based on selection of the one candidate image.