Conventional ultrasound imaging technology used the amplitude information in the ultrasonic linear signal to form grayscale images. Doctors identified the tissues in grayscale images by their structures. Due to the requirement of specific positions and orientations of the ultrasonic probe, this method relied heavily on the doctor's experience and resulted in a high misdiagnosis rate. Compared with ultrasound linear signal, the nonlinear signal contained more information such as frequency and phase that can be used in ultrasound imaging. So we proposed a spectral-based color ultrasound imaging technique based on nonlinear vibration in this paper. First, we used wavelet transform to analyze spectral features. Then, the spectral features were analyzed by principal component analysis to reduce the dimension. Finally, different tissues with different spectral features were color-coded according to the projected coordinates in the eigen-space formed by the principal components, thereby realizing the use of feature colors to distinguish various biological tissues. The experimental results demonstrated that different tissues could be recognized clearly by feature colors, which verified the feasibility of this technique. This technique pioneers the way to ultrasound diagnosis without the necessity of specific positions or orientations of ultrasound probes, reducing the complexity and promoting the efficiency significantly.