Using the power spectra of images and noise for portal imaging systems characterization

© Springer International Publishing AG 2017. In this work an ensemble of portal images and noise images are wavelet transformed. The power distribution of the resulting coefficients is then calculated as a function of scale. Assuming statistical independence between noise and image, the noise power is subtracted to the image power, and the resulting power spectrum (averaged over orien-tations) is found to behave approximately as f¿¿, ¿ = 3.2 ± 0.1, where f is the resolution (inverse of the scale). The power spectrum is calculated in the wavelet domain using a two-dimensional separable transform, with a Haar wavelet basis. No test objects are used in this determination. Clinical portal images and uniform images are the input to the method. In this way, the portal image system performance is studied on an ensemble of images with the same characteristics (contrast, noise and frequency content) as the images used in the clinical procedures.