Integer-based pixel blurriness measure and its applications

In this paper, we propose an integer-based pixel blurriness measure scheme. All values are formed as combination of an integer and a fraction. Three fixed-point operations are implemented, including fractional addition, fractional multiplication, and fractional division. Moreover, mean and variance are computed based on fixed-point operations. Pixel blurriness is defined as slope of pixels' difference over the captured image and four smoothed images, and it is capable of determining proper focus parameters for autofocus. The experiment results will demonstrate that our scheme has low precision loss of pixel blurriness, determines the best focus parameters, and achieves multiple applications.