Selecting Parameters for Image Processing Algorithms: A Case Study Using Retinal Image Segmentation

A challenge in applying image processing algorithms to scientific data is the variation within and across images in a data set. This makes it difficult to select parameters for the algorithms, especially when the data set is large and the variation is unknown. Using the task of segmentation of retinal images, we discuss the challenges that arise in selecting parameters for even a relatively simple algorithm. We show that the availability of ground truth results could make the identification of parameters subjective and propose a simple idea that could benefit the selection of a single parameter.