Mindfulness and relaxation techniques for mental health are increasingly being explored in the human-computer interaction community. Physiological signals and their visualization have often been exploited together in a form of biofeedback with other intervention methods. Here, we aim to contribute to the body of existing work on biofeedback interfaces for mindfulness, with a particular focus on incorporating artistic effects into physiological signal visualization. With an implemented artistic biofeedback interface, we conduct a pilot study where 10 participants attend stress-induction sessions followed by two biofeedback mindfulness sessions: classic biofeedback and artistic visualization. The result demonstrates that artistic visualization-driven biofeedback significantly improves the effectiveness of biofeedback in helping users feel relaxed in comparison with a classic graphical form of biofeedback. Also, it shows that the artistic effect makes it easy to understand what biofeedback represents. Future work includes exploring how advanced physiological computing methods can improve its efficiency and performance.