This publication describes techniques and processes for using image-processing settings (e.g., Auto-Exposure (AE), Auto-Focus (AF), and/or Auto-White Balance (AWB)) to determine an optimal operating point for object detection by an object detector on an imaging device. An operating point is provided to the object detector by a manufacturer to enable the object detector to execute object detection. Through object detection, the object detector determines if an object is identified in the scene based on a confidence score. The optimal operating point has a computed image-processing setting that is closest to an ideal value of the image-processing setting. In an example, a fixed penalty function allows an optimal operating point to be determined using computed AE results for the image at different operating points compared to an ideal AE for the image. The smallest difference between the computed AEs and ideal AE corresponds to the optimal operating point for the image. The process can be repeated for many images to determine an optimal operating point across many types of images. Additionally, the process can be conducted with other image-processing settings, such as AF and AWB, to guide the selection of an optimal operating point across many settings. The determined optimal operating point can be provided to an object detector on an imaging device to provide a positive user experience with the imaging device.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Yang, Ruiduo; Xue, Tianfan; Barron, Jonathan T.; and He, Qiurui, "Using Image-Processing Settings to Determine an Optimal Operating Point for Object Detection on Imaging Devices", Technical Disclosure Commons, (March 04, 2020)