A novel method for direct estimation of axial, lateral and shear strains in tissue using high frame rate ultrasound images is developed. Conventionally, strains are calculated from the spatial derivatives of displacements in different directions, which would result noise amplification and therefore low accuracy of the estimation. The new method uses the local properties of autocorrelation functions to estimate strains directly as parameters of an affine transformation. Unlike other methods in strain measurement, the assumption of isotropic material is not included in the proposed method. The components of strain tensor are estimated independently but concurrently. The spatial constraints of soft tissue on strain tensor is imposed implicitly through spatial autocorrelation of local ultrasonic images. This spatiotemporally resolved measurement enable us to perform principal decomposition to achieve 2-D deformation mode. Error analysis using synthesized ultrasound images show that the accuracy of strain measurement can achieve up to 0.1% for all components. The proposed method is particularly suitable for in-situ measurement of anisotropic soft tissue such as vascular endothelial lining and applicable to use as a point-of-care technique to detect cardiovascular disorders.