Practical noise assessment method

Noise assessment is important to image quality evaluation. It is also important to the criteria for terminating an iterative noise reduction process. To make a meaningful assessment among the distracting features throughout the images, we need to find a strategy to separate the pixels in featureless regions from the rest of the image. We proposed a concept of pseudo-standard deviation (PSD). We take the histogram of PSD with fine bins and calculate the moving average of the histogram. The window length of the moving average is obtained from a monotonical function of the maximum histogram peak location (PSD location). The first peak in filtered histogram gives the most representative noise measure as a desired approximation of true standard deviation. The calculated results compare well with the manually measured standard deviation values. Our method is simple and reliable. It is fully automatic, no need for user to specify the regions of interest. It is suited for batch processing or iterative noise reduction processes.