Image Processing Apparatus Moving Subject Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing techniques struggle to accurately detect moving subject regions, especially in images with high random noise, leading to incorrect detection and handling of stationary or moving regions.

Innovation Solution

An image processing apparatus that includes a likelihood generation unit to detect motion and generate moving subject likelihood, a similarity detection unit to assess similarity between target and peripheral regions, and a correction unit to refine the moving subject likelihood based on detected similarity, thereby improving detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If image combination processing is performed to reduce random noise, then random noise in the combined image is reduced, but moving subjects may appear as multiple images

Engineering Contradiction:

Improverandom noise reductionVSAvoidmoving subject image accuracy

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The image is divided into moving subject regions and stationary regions based on motion detection. Different processing strategies are applied to each region: combination processing is applied to stationary regions for noise reduction, while non-combination processing is applied to moving subject regions to prevent multiple images.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different image processing qualities are applied to different regions of the image. Stationary regions receive combination processing with higher noise reduction, while moving subject regions receive non-combination processing to maintain temporal accuracy and prevent ghosting effects.

Inventive Principle:

Principle #3Local quality

2Difficulty of detecting and measuring

If difference absolute value between images is used to detect moving subject region, then moving subject detection is performed, but random noise increases detection errors

Engineering Contradiction:

Improvemoving subject detection capabilityVSAvoiddetection accuracy

Core Design Contradiction:

Difficulty of detecting and measuringVSMeasurement precision

Solution Approach 1:

A similarity detection unit is introduced as an intermediary to assess the similarity between target regions and peripheral regions. This intermediary provides additional contextual information that helps distinguish true moving subjects from random noise, improving detection accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system uses feedback from similarity detection results to correct moving subject likelihood values. Regions with low similarity to peripheral areas (indicative of true moving subjects) have their likelihood values enhanced, while regions with high similarity (likely noise) have their values suppressed.

Inventive Principle:

Principle #23Feedback

3Productivity

If reduced images are used for moving subject detection, then computational load is reduced, but detection accuracy decreases due to loss of detail

Engineering Contradiction:

Improveprocessing speedVSAvoiddetection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs motion detection on reduced images to save computational resources, then applies similarity detection and likelihood correction on the original full-resolution images for the detected moving subject regions. This partial application of full processing ensures accuracy where needed while maintaining efficiency overall.

Inventive Principle:

Principle #16Partial or excessive action

4Measurement precision

If moving subject likelihood is corrected based on peripheral region similarity, then detection accuracy is improved, but device complexity increases

Engineering Contradiction:

Improvemoving subject detection accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The processing is segmented into distinct functional units: motion detection unit, similarity detection unit, and likelihood correction unit. Each unit handles a specific aspect of the detection process, making the overall complex system more manageable and modular.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS10715729B2Image processing apparatus for detecting moving subject, image processing method, and storage medium

Publication Date: 2020.07.14 CANON KK

AI summary

An image processing apparatus configured to detect a moving subject region includes a likelihood generation unit, a similarity detection unit, and a correction unit. The likelihood generation unit detects a motion of a region in an image and, based on at least two input images, generates a moving subject likelihood for each region. The similarity detection unit detects a similarity between a target region and a peripheral region of the target region for at least one of the input images. The correction unit corrects the moving subject likelihood of the target region based on the detected similarity and the generated moving subject likelihood of the peripheral region. The moving subject region is detected based on the corrected moving subject likelihood.