Different state-of-the-art object detection methods have been applied in agriculture for precision livestock farming. However, the quality and accuracy of livestock such as cow being detected in an image during computer vision tasks depend on the segmentation and extraction techniques used. Mask R-CNN, an instance segmentation method popular for its class and mask regression has been widely applied for cow image segmentation tasks. However, its algorithm relies on simultaneous localization and mapping algorithms, thereby discrediting its ability to completely segment an image foreground from the image background. In this paper, a Mask R-CNN segmentation method integrated with Grabcut is proposed. The method is for detection and complete extraction of an image foreground from the image background. We performed an experiment using edge detection to compare the segmentation that is region-based with the boundary estimation. After comparing the segmentation, we have concluded that the Grabcut integrated with the Mask R-CNN is based on the computation of the edge detection approach employed by our proposed method.