Validation of correction methods for the non-linear response of digital radiography systems

Standard deviation is used as a measure of noise in digital images for quality control. For systems that have a non-linear response between pixel value and air kerma, the standard deviation of the pixel values will not be meaningful unless the image data is linearized using the response function of the system. Two methods are compared for estimating standard deviation without linearizing the image data. Method A estimates the true standard deviation by subtracting the corrected mean pixel value from the corrected sum of the mean pixel value and pixel standard deviation. Method B divides the pixel standard deviation by the point gradient of the response curve. The methods were tested using a range of digital radiographical systems with different response functions between pixel values and air kerma. The aim of this work was to measure the standard deviation and compare the accuracy of the two methods over a wide range of exposures. Both methods are easy to apply and require only basic software. Method A is considered satisfactory for routine quality assurance, as the error is generally <1%; however, the errors increase with reducing air kerma. Method B normally produces errors <0.2%. At low air kerma, the measured errors are up to 8% and 2% for Methods A and B, respectively. Method B is robust over a wide range of air kerma and is easily implemented.