Many methods of measuring contrast-to-noise ratio (CNR) in magnetic resonance imaging (MRI) have been proposed. However, it is not clear which method is best for evaluating clinical or phantom images. In this study we examined the characteristics of the methods of evaluation proposed in the past, and we proposed new CNR evaluation method that improved noise evaluation. We examined the relationship of theoretical CNR value and measurement value when measurement sensitivity was changed. We measured the relationship between number of signal averaged (NSA) and value of CNR. The CNR value changed greatly according to where noise was measured. The measuring method that we proposed in this study was superior for the following reasons: the measurement point of noise and signal are the same; the influence of the low frequency element is slight; and the correlation of measurements and theoretical value is high. The method that we proposed in this study is useful for evaluating phantom images.