Models for biomedical image reconstruction based on integral approximation methods

The most common image representation method for biomedical image reconstruction uses pixels, and the image is assumed to be constant throughout the pixel. Other methods have also been used. In many reconstruction problems, the measured data is approximated by line integrals through the object. This fact suggests a new class of model representation methods based on classical Newton-Cotes methods of integral approximations. These methods use Lagrange polynomials of one variable, and they can be extended to higher dimensions by blending. In 2D, these methods lead to the pixel model, bilinear interpolation, and higher order models. The bilinear interpolation model has been implemented and shown to be superior to the pixel model.