Ghost Detection in Parallax Images via Relative Difference Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to accurately detect unwanted components like ghost and flare in captured images, especially when capturing moving objects, as they require multiple-time image capturing or are less effective with multiple viewpoint images.
Innovation Solution
An image processing method that acquires multiple parallax images and uses relative difference information to detect unwanted components, allowing for accurate detection without multiple-time image capturing, by utilizing the differences between these images to isolate and remove ghost and flare.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple-time image capturing is performed to detect unwanted components, then detection accuracy is improved, but capturing time increases and productivity decreases
Solution Approach 1:
The patent segments the single image capture into multiple virtual viewpoint images by using a light field camera structure with microlens arrays. This allows the system to process multiple perspectives simultaneously from a single capture event, enabling unwanted component detection through comparative analysis of these segmented viewpoints without requiring multiple sequential captures.
Solution Approach 2:
The patent transitions from temporal dimension (multiple captures over time) to spatial dimension (multiple viewpoints from single capture). By capturing light field information that encodes angular and spatial relationships, the system achieves multi-perspective analysis in a single shot, resolving the contradiction between detection accuracy and capturing speed.
2Device complexity
If only two images (main image and sub image) are used for ghost detection, then processing complexity is reduced, but detection effectiveness decreases when three or more viewpoint images are available
Solution Approach 1:
The patent merges multiple viewpoint images (three or more) into a unified processing framework. By combining the information from all available viewpoints and applying comparative analysis across the entire set rather than pairs, the system achieves enhanced ghost detection effectiveness while maintaining manageable processing complexity through systematic algorithms.
Solution Approach 2:
The patent creates a universal detection framework that works with any number of viewpoint images (two or more). The same processing methodology can be applied regardless of whether two, three, or more images are available, making the system adaptable and maximizing detection effectiveness without requiring separate processing pipelines for different image counts.
Data Source
AI summary
The image processing method includes acquiring multiple parallax images produced by image capturing of an object, the parallax images having a parallax to one another. The method further includes acquiring, by using the respective parallax images as base images, relative difference information on a relative difference between each of the base images and at least one other parallax image in the multiple parallax images, and detecting an unwanted component contained in each of the parallax images by using the relative difference information.