Fruit image segmentation based on a novel watershed algorithm

Previous research about automatic vision systems applied to fruit imaging segmentation fails. Watershed transformation is a powerful morphological tool for image segmentation. However, the performance of segmentation methods based on watershed depends largely on gradient of the image. It's not perfect to segment object with single-scale gradient. In this paper we present a multi-scale algorithm for computing morphological gradient images, after this transformation, there exists a strict causality between gradient watershed under different scales of geodesic reconstruction. The results demonstrate the superior performance of the watershed segmentation over the conventional techniques.