Edge detection algorithms play a vital role in image processing, computer vision, and machine vision. There is a huge demand for efficient edge detection algorithms for identifying the exact region of interest in an image. The nature-inspired metaheuristic algorithms are more promising than traditional algorithms owing to their stochastic characteristics. The concept of the Flower Pollination Algorithm (FPA) has recently gained much attraction due to solving several complicated optimization problems. In this study, an efficient edge detection algorithm has been developed using FPA for identifying edges in an image. The proposed FPA is capable of identifying the edges with minimal parameterized values, and the parameters are initialized automatically using the concept of maximum entropy and polynomial curve fitting distribution. The performance of flower pollination based edge detection algorithm on ground truth images was compared with other existing methods like Sobel, Canny, Ant Colony based edge detection method, PSO method, and fuzzy-genetic based edge detection methods using Receiver Operating Characteristics (ROC) curve and Area Under Curve (AUC) method. The proposed method performed better compared with other existing methods and had significantly high ROC and AUC indices. Also, the performance of the proposed algorithm was tested on real-time banana leaf disease images and compared with other existing methods in terms of the Shannon entropy index. The segmented images of banana leaf disease through enhanced FPA showed significantly lower mean entropy value, indicating extra-ordinary accuracy and negligible uncertainty of the enhanced FPA. The proposed FPA seems to be promising in developing image processing modules for crop disease diagnosis.