Ghost-Free HDRI Creation Using Probabilistic Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for creating High Dynamic Range Images (HDRI) from Low Dynamic Range Images (LDRIs) acquired with a moving camera suffer from the 'ghost phenomenon' due to object motion, where objects appear multiple times in the final image, making it difficult to accurately identify and exclude ghost areas, leading to undesired ghosting in the final HDRI.
Innovation Solution
An apparatus and method that collect LDRIs while varying exposure, identify a reference frame based on visibility, calculate a joint probability function, and use a global transfer function to create an HDRI, effectively preventing ghosting by defining ghost areas probabilistically rather than through direct detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional ghost detection schemes are used to identify and exclude ghost areas, then ghosting can be partially reduced, but measurement precision of ghost area identification deteriorates due to difficulty in exactly and accurately identifying ghost areas
Solution Approach 1:
The patent transforms the ghost detection problem from spatial domain to probability domain by changing the parameter representation. Instead of directly detecting ghost areas in image space, the method computes joint probability functions P(I_ref, I_n) that characterize the relationship between reference frame intensities and other frame intensities. This parameter transformation enables accurate ghost area identification by evaluating probability deviations rather than relying on direct visual detection.
Solution Approach 2:
The patent introduces joint probability functions as an intermediary mechanism between the reference frame and other frames. Rather than directly comparing image pixels to identify ghosts, the method uses probability functions P(I_ref, I_n) as a mediator to indirectly detect ghost areas. This intermediary approach allows for more precise ghost identification by capturing statistical relationships that are not apparent through direct pixel comparison.
2Device complexity
If direct detection methods are used to identify ghost areas, then detection complexity is reduced, but manufacturing precision of ghost-free HDRI deteriorates due to incorrect identification of ghost areas
Solution Approach 1:
The patent performs preliminary computation of joint probability functions and histograms before final ghost detection. By pre-calculating P(I_ref, I_n) and the joint histogram H(I_ref, I_n) from collected frames, the method prepares probability-based detection data in advance. This preliminary action enables more accurate ghost identification during the HDRI creation process without increasing real-time detection complexity.
3Illumination intensity
If multiple LDRIs are collected with varying exposure to create HDRI, then image quality and dynamic range are improved, but ghost phenomenon worsens due to object motion during continuous capture
Solution Approach 1:
The patent dynamically adapts the ghost detection process to handle motion by using probability-based methods that are insensitive to temporal changes. Instead of assuming static scenes, the method computes joint probability functions that naturally accommodate object motion by evaluating the statistical relationship between frames. This dynamic approach allows multiple exposures to be combined effectively even when objects move during capture.
Data Source
AI summary
An apparatus and method creating a ghost-free High Dynamic Range Image (HDRI) based on filtering are provided. It is possible to effectively prevent a ghost phenomenon from occurring when a single HDRI is created from a plurality of LDRIs, by defining a ghost area using a probability based on a global transfer function indicating a relationship for intensities of several frames, rather than searching for or identifying a ghost area in a single or each image.