HDR Image Composition Using Motion Vector Occlusion Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating HDR images from multiple exposure images fail to effectively handle moving objects, resulting in multiple overlapping images, unnatural blur, and afterimages due to neglecting occlusion characteristics.
Innovation Solution
An image processing apparatus that acquires multiple images with different exposures and determines the direction of object movement to compose HDR images by selecting appropriate pixels based on occlusion characteristics, using techniques like optical flow and difference image calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If HDR composition is performed by capturing multiple images with different exposure amounts, then the dynamic range of the captured scene is improved, but multiple overlapping images and unnatural blur occur when objects move during capturing
Solution Approach 1:
The patent calculates movement vectors between images in advance before performing HDR composition. By determining the movement amount and direction of objects beforehand, the system can pre-identify occlusion areas and selectively use pixels from appropriate images, preventing multiple overlapping images and unnatural blur in the final HDR image
Solution Approach 2:
The patent divides the image into different regions based on movement characteristics. By segmenting the image into moving object areas and stationary areas, the system applies different composition strategies to each region, using pixels from images that best represent each area while avoiding occlusion artifacts
2Ease of operation
If existing HDR composition methods are used without considering occlusion characteristics, then the processing is simple, but disappearance areas and appearance areas cause multiple overlapping images and afterimages
Solution Approach 1:
The patent performs preliminary calculation of movement vectors and identification of occlusion characteristics before HDR composition. By determining which areas are disappearance areas or appearance areas in advance, the system can pre-select appropriate pixels from different exposure images, eliminating the need for complex post-processing while avoiding occlusion artifacts
Solution Approach 2:
The patent applies different composition strategies to different regions of the image based on local occlusion characteristics. By identifying disappearance areas and appearance areas locally, the system selects pixels from appropriate images for each region, ensuring high composition accuracy in moving object areas while maintaining processing efficiency
Data Source
AI summary
The image processing apparatus according to one embodiment of the present invention has: an acquisition unit configured to acquire a first image captured with a first amount of exposure, a second image captured with a second amount of exposure different from the first amount of exposure before the first image is captured, and a third image captured with the second amount of exposure after the first image is captured; a specification unit configured to specify a direction of movement of an object by using at least two images of the first image, the second image, and the third image; and a composition unit configured to generate an HDR image by composing the first image with one of the second image and the third image for each pixel in accordance with the direction of movement of the object.