A Generic Parameter Optimization Workflow for Camera Control Algorithms

Cameras are often controlled by algorithms adapting the image capturing process parameters (like exposure or gain) to the present scene. In most cases these algorithms have to be parametrized by a parameter set in order to define the behavior of the control. The issue of selecting the best parameter set for a specific application or environment arises. The parameter selection is not a simple task since the internal structure of the control algorithm is often not sufficiently known by the user. Only the inputs (parameter sets) can be specified and the outputs (images) can be analyzed. This paper presents a generic workflow for the determination of a parameter set which achieves good image quality for a chosen application with specific light conditions. The developed workflow is able to deal with any control algorithm and any chosen application. In general, the four main steps of the developed workflow are: 1. Build a database of images with their related parameter sets, 2. Evaluate which image criteria are best to assess the image quality for the particular application, 3. Choose an optimization method, 4. Optimize the parameter sets. The presented workflow is developed and examined based on the example of real-world automotive scenarios. At the end of the paper experimental results confirm that the optimized camera parameters achieve a meaningful and useful optimization result regarding the images captured by the camera.