Application of Image Filtering Technique Based on Bounded Mean Oscillation Model in Studying on Rice Grain Morphology

Abstract

This study constructed a bounded mean oscillation (BMO) filter via the BMO algorithm and anisotropic nonlinear partial differential equation (PDE) to both denoise and enhance the digital image of rice grains. The Perona–Malik PDE model was used as control filter. Based on the quantitative evaluation of the morphological characteristics of rice grains, as obtained from preprocessed images, the BMO filtering effect is discussed. The results showed that grain length, grain width, and the length–width ratio obtained from BMO filter processed images did not significantly differ from manual measurements ([Formula: see text]). Moreover, a strong positive correlation was found between the average grain area and the thousand grain weight ([Formula: see text], [Formula: see text]). The BMO filter was less disturbed by noise and the structure of the utilized algorithm was simpler compared with the Perona–Malik filter. The developed BMO filter was also superior to the Perona–Malik filter in retaining fine edge features of digital images. Moreover, its filtering effect remained stable for grain images of different rice varieties.