https://doi.org/10.20965/jaciii.2020.p0963 Cell Image Segmentation via Improved Markov Random Field Paper: Robust Cell Image Segmentation via Improved Markov Random Field Based on a Chinese Restaurant Process Model Dongming Li∗1,∗2, Changming Sun∗3,SuWei∗4,YueYu∗2,∗5, and Jinhua Yang∗1,† ∗1College of Opto-Electronic Engineering, Changchun University of Science and Technology No.7089 Weixin Road, Chaoyang District, Changchun, Jilin 130022, China E-mail:
[email protected] ∗2School of Information Technology, Jilin Agricultural University No.2888 Xincheng Road, Jingyue District, Changchun, Jilin 130118, China ∗3CSIRO Data61 P.O. Box 76, Epping, New South Wales 1710, Australia ∗4Modern Educational Technology Center, Changchun University of Chinese Medicine No.1035 Boshuo Road, Jingyue District, Changchun, Jilin 130117, China ∗5College of Artificial Intelligence, Tourism College of Changchun University Sheling Town, University Campus District, Changchun, Jilin 130607, China †Corresponding author [Received October 5, 2020; accepted November 18, 2020] In this paper, a segmentation method for cell images common approaches for cell segmentation is image inten- using Markov random field (MRF) based on a Chi- sity thresholding which suffers from the issue of inten- nese restaurant process model (CRPM) is proposed. sity inhomogeneity. Watershed based methods are com- Firstly, we carry out the preprocessing on the cell im- monly utilized for clustered nuclei segmentation. Dunn ages, and then we focus on cell image segmentation proposed the fuzzy C-means (FCM) method for image using MRF based on a CRPM under a maximum a segmentation [2]. The FCM method is an unsupervised posteriori (MAP) criterion. The CRPM can be used clustering algorithm, which has been widely used in the to estimate the number of clusters in advance, adjust- field of image segmentation.