Night Image Enhancement via Multi-Exposure Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pedestrian detection systems in vehicles, relying on thermal or near-infrared cameras, increase vehicle costs and face challenges in forming clear night images due to underexposed and overexposed zones, making it difficult for drivers to identify pedestrians and other important elements.
Innovation Solution
A system and method that utilizes a front camera to acquire images with varying sensitivities and exposure times, combined with a map of acquisition parameters, to enhance image processing for driver assistance systems, including automatic headlight control and zone-of-interest definition, which creates a composite image to highlight important elements while reducing saturation and enhancing contrast.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If thermal camera or near infrared camera is used for pedestrian detection, then detection capability is improved, but vehicle cost increases due to additional sensors
Solution Approach 1:
The patent uses a standard front camera to capture visible light images and creates a processed version that simulates the effect of infrared imaging through image processing algorithms. This copying approach allows the system to achieve pedestrian detection capabilities without requiring actual infrared sensors, thereby reducing vehicle cost while maintaining detection reliability
Solution Approach 2:
The patent replaces the need for additional physical sensors (thermal or infrared cameras) with image processing software that runs on existing camera hardware. This substitution of mechanical/system complexity with computational processing achieves the same functional goal at lower cost
2Device complexity
If front camera is used for night image acquisition, then cost is reduced, but image quality deteriorates due to underexposed and overexposed zones
Solution Approach 1:
The patent divides the night scene into multiple zones with different exposure requirements and captures separate images for each zone using the same camera at different exposure settings. This segmentation allows each region to be optimally exposed while maintaining overall image quality without requiring additional sensors
Solution Approach 2:
The patent changes the exposure parameter of the camera to capture multiple images at different exposure levels. By varying this single parameter, the system obtains both underexposed and overexposed images that can be combined to create a high-quality night image with proper detail in both bright and dark areas
3Manufacturing precision
If multiple images with different sensitivities are acquired, then image processing performance is improved, but acquisition time increases
Solution Approach 1:
The patent uses periodic switching between different exposure settings to capture multiple images rapidly in sequence. This periodic action allows the camera to cycle through different sensitivity levels quickly, acquiring the necessary variety of images for processing while minimizing total acquisition time
Solution Approach 2:
The patent pre-plans the sequence of exposure settings to be used and prepares the image processing pipeline in advance. This preliminary preparation allows the system to efficiently capture and process multiple images with different sensitivities without time loss, as the processing steps are already configured before image acquisition begins
Data Source
Figure 1~2
AI summary
A system for forming nighttime images for a motor vehicle provided with headlights (2a), a camera (3), a device (7) for displaying the images formed and at least one driver assistance system capable of transmitting an area of interest included in at least one image acquired by the camera (3). The system comprises: a map (3a) of the acquisition parameters of each of the driver assistance systems connected to the camera (3), a switch (3b) capable of memorising the images acquired by the camera (3) and transmitting them to the various driver assistance systems, a calculation means (5) capable of determining a current area of interest from the combination of the areas of interest, an image processing means (6) capable of determining at least one image in which the intensity of each pixel is equal to the intensity of the corresponding pixel in the designated image in the current area of interest.