An improved active contour model for food image segmentation

Abstract

Active contour (snake) segmentation is extensively used in image processing and analysis applications, particularly for identifying object boundaries. This method is adopted for food image segmentation, wherein the boundaries of the foods in an image are the objects of interest. In this research, a region-based active contour, which is regarded as an energy-minimizing process, is used for image segmentation. Additionally, a modified active contour method is presented in this paper using the artificial bee colony (ABC) algorithm to optimize the weights of the external energy function in the original active contour (AC) method. A stopping criterion is also introduced to the AC method, wherein the snake movement of the contour will cease after a certain number of unimproved snake movements. The food image dataset was collected manually for this research; it comprises 102 images of different food types and positions within each image. The modified active contour method demonstrated significant improvements in fewer iterations and segmentation quality compared with the original method.