This paper describes a new method of non-invasive temperature measurement using ultrasound B-mode images. In the clinic, a real-time monitoring system to acquire the temperature distributions around the targeted area is necessary for effective hyperthermia treatments. We proposed the temperature measurement method by image processing techniques. The method of measuring temperature distributions was based on the thermal dependence of local change in speed of sound and thermal expansion. In this paper, we described the method to estimate temperature distributions inside agar phantom and porcine liver heated by needle-electrode applicator. First, we described the principle of temperature measurement from ultrasound images by image processing techniques. Second, we described the theory of template matching technique and sub-pixel processing. Third, we showed the algorithm of the developed temperature measurement method based on ultrasound image processing technique. Fourth, we estimated tissue dependence of thermal coefficient ktissue which was used in temperature calibration. Finally, we estimated temperature distributions inside agar phantom and porcine liver tissue heated by needle applicators. From our heating experiments, it was confirmed that the proposed method could estimate temperature distributions inside the heated object non-invasively. The measured temperature from ultrasound images of the heated porcine liver tissue was close to the result of fiber optical thermometer within an error of 1°C or less. From these results, it was confirmed that the proposed method is useful for effective hyperthermia treatments.