DNN Object Detection for Real-Time Image Sensor Parameter Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image adjustment techniques in consumer devices, such as web cameras, are not feasible for real-time application due to resource constraints and lack object-specific processing, resulting in suboptimal image quality for detected objects.
Innovation Solution
A method using a deep neural network to detect objects in images and adjust image sensor parameters based on object-specific statistical information, generating weighted image values to optimize image acquisition for detected objects, such as faces, improving brightness and exposure settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If broad post-processing image adjustment techniques are applied to the entire image, then overall image quality may be improved, but processing resources required become too high for real-time application in consumer devices
Solution Approach 1:
The patent segments the image into multiple regions and identifies specific objects of interest within those regions. By focusing image adjustment techniques only on regions containing detected objects rather than processing the entire image, the system achieves high-quality object-specific enhancement while reducing overall processing requirements to enable real-time performance in consumer devices.
Solution Approach 2:
The patent applies different image adjustment parameters to different regions of the image based on detected objects. Instead of uniform global adjustment, the system calculates object-specific statistical information and applies tailored brightness, contrast, and other parameters only to regions containing detected objects, optimizing image quality where it matters most while conserving processing resources.
2Manufacturing precision
If broad post-processing image adjustment techniques are applied to the entire image, then some image parameters may be enhanced, but image quality of specific objects may be reduced or not benefited
Solution Approach 1:
The patent calculates object-specific statistical information for detected objects and generates weighted adjustment parameters tailored to each object's characteristics. This enables the system to optimize image quality for each detected object individually, adapting parameters such as brightness and contrast to the specific needs of each object rather than applying uniform adjustments that may degrade certain objects.
Solution Approach 2:
The system uses detected object information and object-specific statistical information as feedback to dynamically adjust image parameters. By continuously analyzing object characteristics and adjusting parameters based on this feedback, the system adapts to different objects and scenes, ensuring optimal image quality for each detected object while maintaining versatility across various imaging conditions.
Data Source
AI summary
Systems and methods directed to adjusting an image based on a detected object depicted in the image are described. The method may include receiving an image from an image sensor, receiving statistical information associated with the image, detecting an object depicted in the image using a deep neural network, identifying object-specific statistical information for the detected object, generating a weighted object-specific parameter based on the object-specific statistical information, generating a weighted-image value based on the weighted object-specific parameter, providing the weighted-image value to the image sensor, where the image sensor is configured to update one or more image sensor parameters based on the weighted-image value, and acquiring an image from the image sensor updated with the one or more image sensor parameters.