Many ubiquitous computing projects have addressed health and wellness behaviors such as healthy eating. Healthy meal recommendations have the potential to help individuals prevent or manage conditions such as diabetes and obesity. However, learning people's food preferences and making healthy recommendations that appeal to their palate is challenging. Existing approaches either only learn high-level preferences or require a prolonged learning period. We propose Yum-me, a personalized healthy meal recommender system designed to meet individuals' health goals, dietary restrictions, and fine-grained food preferences. Marrying ideas from user preference learning and healthy eating promotion, Yum-me enables a simple and accurate food preference profiling procedure via an image-based online learning framework, and projects the learned profile into the domain of healthy food options to find ones that will appeal to the user. We present the design and implementation of Yum-me, and further discuss the most critical component of it: FoodDist, a state-of-the-art food image analysis model. We demonstrate FoodDist's superior performance through careful benchmarking, and discuss its applicability across a wide array of dietary applications. We validate the feasibility and effectiveness of Yum-me through a 60-person user study, in which Yum-me improves the recommendation acceptance rate by 42.63% over the traditional food preference survey.