An Ultrasound Image Quality Improvement Approach Based on Non-Subsampled Shearlet Transform and Maximum Local Variation-Based Unsharp Masking

Speckle noise is the most prominent source of poor ultrasound image quality. Likewise, the contrast in an ultrasound image is often low. Therefore, this paper presents an approach for improving ultrasound image quality. Non-subsampled Shearlet Transform (NSST) was employed to reduce speckle noise, while Maximum Local Variation-Based Unsharp Masking (MLVUM) was used to increase image contrast and sharpness. The number of decomposition levels and shearing directions in each decomposition level of the NSST was varied. The number of decomposition levels varies by 2, 3, and 4, while the number of shearing directions varies by 4, 8, and 16. The effectiveness of NSST in reducing speckle noise was evaluated using PSNR, EPI, ENL, and SSI. Meanwhile, ALC and AG were used to measure the difference in contrast and sharpness before and after using MLVUM. The results indicate that NSST has a good despeckling ability. Speckle noise gradually decreases as the number of image decomposition levels increases. NSST with 4-level decomposition provides the best speckle suppression. Meanwhile, varying the number of shearing directions at each decomposition level does not have a significant effect on denoising results. Furthermore, the use of MLVUM after the denoising step enhances the sharpness and contrast of the denoised images.