Ghost Detection in Parallax Images via Multi-Stage Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to effectively separate parallax components and unnecessary components, such as ghosts, in shot images without requiring multiple image pickups, especially when parallax components are significant, leading to degraded detection efficiency.
Innovation Solution
An image processing method that determines a first unnecessary component from parallax images based on relative difference information, generates a blurred image, creates a mask, and then determines a second unnecessary component by reducing the parallax component from the first unnecessary component, allowing for the separation of parallax and unnecessary components without multiple image pickups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If parallax images are used to detect ghosts, then ghost detection is applicable to still image and moving image pickup of moving objects, but the ghost detection effect is degraded when using parallax images from three viewpoints or more
Solution Approach 1:
The patent segments the ghost detection process into multiple stages: first detecting ghosts using parallax images, then performing secondary detection to eliminate false positives. This segmentation allows the system to maintain high detection precision while processing images from multiple viewpoints (three or more), resolving the contradiction between versatility and detection accuracy.
Solution Approach 2:
The patent introduces an intermediary verification step using multiple parallax image comparisons. Instead of directly detecting ghosts from three or more viewpoints, the system uses an intermediate detection phase to identify candidate regions, then verifies them through additional parallax analysis, thereby maintaining detection precision across multiple viewpoints.
2Measurement precision
If parallax component correction is applied before ghost detection, then focused object regions can be processed, but blurred object regions require detailed distance information making the process complex
Solution Approach 1:
The patent applies partial correction of the parallax component - only correcting regions that are sufficiently focused while leaving blurred regions uncorrected. This partial action approach allows the system to achieve good ghost detection in focused areas without incurring the high complexity of full parallax correction across the entire image, including blurred regions.
Solution Approach 2:
The patent implements local quality processing by applying different treatments to different image regions: focused regions receive parallax correction and ghost detection, while blurred regions are handled differently based on their distance information. This local differentiation reduces overall process complexity while maintaining detection accuracy where needed.
3Device complexity
If ghost detection is performed without correcting parallax difference, then processing is simpler, but the parallax component and ghost component cannot be separated
Solution Approach 1:
The patent segments the detection signal into parallax component and ghost component through multi-stage analysis. By comparing parallax images from multiple viewpoints and analyzing the characteristics of detected components, the system separates these overlapping signals without requiring complex pre-correction processing, thus maintaining both simplicity and accuracy.
Solution Approach 2:
The patent uses feedback mechanisms where the detected components are analyzed and used to refine subsequent detection steps. The system detects initial candidates, analyzes their characteristics to distinguish parallax from ghosts, and uses this feedback to improve separation accuracy in later stages without adding significant processing complexity.
4Object-affected harmful factors
If low-pass filter is applied to remove parallax component, then distant views can be processed, but the ghost signal also decreases making detection difficult
Solution Approach 1:
The patent applies local quality filtering by using different filtering strengths in different image regions. In distant view areas where parallax is minimal, stronger low-pass filtering is applied to remove parallax components. In closer regions where ghost signals are stronger, weaker filtering is used to preserve the ghost detection signal, thus balancing parallax removal with ghost signal preservation.
Solution Approach 2:
The patent dynamically changes the filtering parameters based on local image characteristics. The low-pass filter's cutoff frequency and strength are adjusted according to the detected parallax magnitude and expected ghost signal strength in each region, allowing optimal balance between parallax removal and ghost detection across different scene depths.
Data Source
AI summary
An image processing method includes the steps of determining a first unnecessary component contained in each of a plurality of parallax images based on a plurality of pieces of relative difference information of the parallax images, generating a first image by reducing the first unnecessary component from the parallax images, generating a blurred image by adding blur to the first image, creating a mask based on the blurred image and the first image, and determining a second unnecessary component based on the first unnecessary component and the mask.