The invention relates to a high-resolution minimum variance ultrasonic imaging method based on frequency domain segmentation and belongs to the acoustic imaging field. The method comprises the following steps of firstly, carrying out delay processing on a sampling signal received by an array element to obtain echo data required by ultrasonic focusing; secondly, selecting a frequency domain segmentation optimal window function according to a maximum concentration measurement criterion of an adaptive window function in STFT; converting an ultrasonic echo signal into a narrow-band sub-signal by using STFT; carrying out conjugate symmetry processing on a first half of narrow-band sub-signal to generate the other half of narrow-band signal by using conjugate symmetry; successively dividing a receiving array into sub-arrays with overlapped array elements, and carrying out forward and backward smoothing and diagonal loading processing on a frequency domain signal to obtain a sample covariancematrix; and finally, processing a frequency domain segmented minimum variance beam forming weight by using inverse fast Fourier transform to obtain a final time domain adaptive beam forming signal. By using the method, an ultrasonic imaging resolution can be remarkably improved, a contrast ratio is increased, and ultrasonic imaging quality can be integrally improved.