Background: Research on rice (Oryza sativa) roots requires the automatic analysis of root structure using image processing. It is challenging for a digital filter to identify the roots from the obscured and cluttered background, and to separate the primary roots from lateral roots. The original Frangi filter (FF), presented by Alejandro F. Frangi in 1998, is a low-pass filter dedicated to blood vessel image enhancement. Considering the similarity between vessels and roots, the FF is applied to identify the roots. However, the original FF only enhances the tube-like primary roots but erases the lateral roots. Hence, a new method is developed to meet the demands by simultaneously maintaining the primary and lateral morphological structure of roots.
Result: In this work, a crucial part of the FF, Gaussian filtering, is redesigned to discriminate against the primary and lateral roots in a 2-D image. Inspired by the structure-awareness of the FF, an Improved Frangi Filtering Algorithm (IFFA) designed for plant roots is proposed. First, Multilevel image thresholding, connected-components labeling and width correction are used to optimize the resultant binary image. Then, to enhance the local structure, the truncated Gaussian kernel is modified resulting in more discernible lateral roots. Compared to the original FF and the Automatic Root Image Analysis (ARIA), a commercial software, IFFA is a faster and more accurate algorithm achieving an identification accuracy of 97.48%.
Conclusion: IFFA is an effective 2-D filtering approach to enhance the roots of rice (Oryza sativa) for segmentation and further biological research. IFFA is faster than ARIA and the original FF, and IFFA’s accuracy outperforms its counterparts as per the Intersection Over Union (IOU) and Dice Similarity Coefficient (DSC) criteria.