The invention discloses a denoising system for ultrasonic image speckle noise. The denoising system comprises a phase field loading module, a normalization processing module, a partial differential processing module, a free energy noise reduction module, a threshold value detection module and an image display module. The denoising method comprises steps as follows: S1, ultrasonic image loading; S2, normalization processing; S3, partial differential processing; S4, free energy noise reduction; S5, threshold value detection; S6, image display. Denoising is performed according to the statistical rule of minimization of free energy, and through analysis of noise attributes, speckle noise with the average value being zero can be weakened and even counteracted through iterative computation by an established proper nonlinear partial differential equation. In a flat area, the main purpose of denoising is to eliminate scrambled speckles, and an image becomes more uniform. In a boundary area, accurate recognition is firstly made to the boundary, details are reserved, and noise and boundary change are separated in the evolution process.