Scatter Kernel Estimation for Industrial 3D Cone Beam CT

The patent transforms the scatter kernel calculation from a complex integral computation into a parameter-based estimation problem. By representing the scatter kernel through specific parameters (amplitude, width, shape) and using optimization algorithms to determine these parameters, the method achieves both accuracy and efficiency. The scatter kernel is modeled as a function with adjustable parameters that can be optimized to match actual scatter behavior without requiring full numerical integration.

The patent replaces the traditional mechanical/computational approach of numerical integration with an optimization-based approach. Instead of directly computing the scatter kernel through intensive numerical methods, the system uses optimization algorithms (such as least squares fitting or gradient descent) to estimate kernel parameters from measured or simulated data, significantly reducing computational burden while maintaining accuracy.

The patent creates a simplified model (copy) of the scatter kernel that captures its essential characteristics without requiring the full physical complexity. By representing the kernel through parametric equations and optimizing these parameters, the method produces an accurate approximation without needing additional physical measurements or complex experimental setups. The parametric model serves as a computationally efficient copy of the true scatter kernel.

The optimization algorithm automatically determines the scatter kernel parameters by fitting the model to available data, eliminating the need for manual calibration or additional physical operations. The system self-adjusts the kernel parameters to minimize the difference between the model prediction and actual scatter behavior, achieving accurate estimation without external intervention or complex physical procedures.