It is not easy to develop a new image processing algorithm by trial and error for each biological target or measurement method. Machine learning can automatically estimates parameters of a model from training data and is useful to develop the image processing algorithm. Especially approaches using deep learning have attracted attention. In this research, we develop a local region classifier using machine learning and a segmentation method of biological images. For that purpose, we develop machine learning algorithm for local region classifier and image segmentation algorithm using the outputs of the local region classifier.