Robust Pre-Processing Module for Leaf Image Analysis

Abstract

Flora on earth is natural reserve for medicines. It has great healing power if encashed and conserved with great faith and devotion. Knowledge about flora empowers the methods for nurturing the medicines. Digital database creation, automation of plant recognition and identification of plant maturity can play a vital role in medicine extraction. The system for automation of process should be robust to handle on-site data so as to make the process less destructive. Pre-processing algorithms can equip the system with accuracy and robustness. The paper proposes the algorithms used for pre-processing the raw image which would aid the feature extraction and classification methods. Adaptive enhancement method for non-uniform illumination equalizes the underexposed and over exposed part in an image. Also methods like deblurring, orientation correction, size normalization in prescribed sequence improves the image quality for the later stages. The case study undertaken considers 38000 images and accuracy achived is of about 98%.