Adaptive Image Thresholding for Tone-Mapped Light Source Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems inaccurately detect vehicle lights due to tone mapping operations, leading to false positives from non-light sources and missed detections of distant light sources, as they compress images to 8-bit formats, losing scene brightness information.
Innovation Solution
A non-linear image transformation is applied to determine light source identification thresholds, using a modified gamma correction technique and an image pyramid to adapt thresholds based on real-world brightness, neighborhood information, and regions of interest, ensuring accurate light source detection in compressed images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If tone mapping compression is applied to reduce dynamic range and save computational resources, then storage and processing efficiency are improved, but scene brightness information is lost leading to inaccurate light source detection
Solution Approach 1:
The patent applies preliminary action by performing non-linear transformation on the compressed image data before light source detection. The system pre-calculates transformation thresholds and applies gamma correction or similar transformations to recover brightness information that was compressed during tone mapping. This allows the system to work with compressed 8-bit images while still achieving accurate light source detection by transforming the data into a space where brightness relationships are better preserved.
Solution Approach 2:
The patent changes parameters by transforming the image data from a compressed brightness space to a transformed brightness space using non-linear functions. The system adjusts transformation parameters such as gamma values and threshold levels to optimize the recovery of scene brightness information. By changing the parameter space through non-linear transformation, the system can accurately distinguish light sources from non-light sources even when working with compressed image data.
2Device complexity
If a global light source detection threshold is applied to the compressed image, then processing is simplified, but distant light sources are missed and false positives from illuminated regions increase
Solution Approach 1:
The patent applies local quality by transitioning from a global threshold approach to a local threshold approach. The system divides the image into multiple regions and applies different detection thresholds to different areas based on local characteristics. By calculating local statistics such as mean and standard deviation of pixel intensities in neighboring regions, the system adapts the detection threshold to local brightness conditions. This allows distant light sources in darker regions to be detected while avoiding false positives from illuminated regions such as lane markings or reflectors.
Solution Approach 2:
The patent implements dynamics by making the detection threshold adaptive rather than static. The system dynamically adjusts thresholds based on local image characteristics, transforming the fixed global threshold into a variable local threshold that responds to changing brightness conditions across different regions of the image. This dynamic adaptation improves detection reliability by matching the threshold to the actual local scene conditions.
3Use of energy by stationary object
If tone mapping is used to compress raw images to 8-bit format, then computational resources are saved, but illuminated non-light sources have intensities comparable to true light sources causing false positives
Solution Approach 1:
The patent introduces an intermediary transformation step between the compressed image data and the light source detection process. The non-linear transformation acts as an intermediary that recovers brightness information from the compressed 8-bit image by applying gamma correction or similar transformations. This intermediary process restores the brightness relationships that were compressed during tone mapping, enabling the system to distinguish between illuminated non-light sources and true light sources based on their transformed intensity values and spatial characteristics.
Data Source
AI summary
A method of identifying a light source in an image, the method includes receiving an image pyramid representing a tone-mapped image by a reference pixel array of low resolution, and by one or more test pixel arrays of higher resolution than the reference pixel array. The method includes comparing pixel values of test pixels in the image with respective brightness thresholds. A light source is identified at a position at which a test pixel value exceeds a respective brightness threshold. The method includes, for each test pixel, defining the respective brightness threshold as a non-linear function of a pixel value of a reference pixel in the reference pixel array which has an area in the image which includes the position of the test pixel and which has a pixel value which is a mean pixel value of the test pixels in the area in the image of the reference pixel.