Open Access Open Access  Restricted Access Subscription Access

De-Speckling of B-Mode Breast Ultrasound Images Using Wavelet Shrinkage Filters:A Comparative Analysis


Affiliations
1 Department of Computer Engineering, Punjabi University, Patiala, India
 

Speckle noise is an inherent yet undesirable residual part of the breast ultrasound images, which significantly demean the visual quality and limits the accuracy of automatic diagnostic techniques. Therefore, speckle elimination is a necessary task before further processing of ultrasonic images. Speckle reduction from breast ultrasound images results in blurring of lesion margins and other sharp details which may carry the significant diagnostic information. Among various denoising methods used in the literature for enhancement of breast ultrasound images, wavelet based techniques are gaining importance, because of their time-frequency and multi-scale analysis. Basic de-noising methods that use the wavelet transform, make use of of three steps - In the first step, it computes the wavelet transform of the noisy input image, the second step is used to apply thresholding in order to remove noise on the detailed coefficients and finally inverse wavelet transform is applied of the modified coefficients to get the denoised image. In this paper, the performance of various wavelet shrinkage techniques is reviewed. An Experimental analysis of wavelet based methods including Visu Shrink, Sure Shrink, Bayes Shrink, and a hybrid method (wavelet shrinkage with guided filter) is carried out. For performance comparison, signal to noise ratio, structural similarity index and edge preservation index, parameters are used. From the experimental results, it is concluded that hybrid filter outperform the traditional wavelet shrinkage methods.

Keywords

Speckle Noise, Breast Ultrasound (BUS) image, Wavelet shrinkage, Signal to Noise Ratio (SNR), Structural Similarity Index (SSIM), Edge Preservation Index (EPI).
User
Notifications
Font Size

Abstract Views: 169

PDF Views: 2




  • De-Speckling of B-Mode Breast Ultrasound Images Using Wavelet Shrinkage Filters:A Comparative Analysis

Abstract Views: 169  |  PDF Views: 2

Authors

Madan Lal
Department of Computer Engineering, Punjabi University, Patiala, India
Lakhwinder Kaur
Department of Computer Engineering, Punjabi University, Patiala, India

Abstract


Speckle noise is an inherent yet undesirable residual part of the breast ultrasound images, which significantly demean the visual quality and limits the accuracy of automatic diagnostic techniques. Therefore, speckle elimination is a necessary task before further processing of ultrasonic images. Speckle reduction from breast ultrasound images results in blurring of lesion margins and other sharp details which may carry the significant diagnostic information. Among various denoising methods used in the literature for enhancement of breast ultrasound images, wavelet based techniques are gaining importance, because of their time-frequency and multi-scale analysis. Basic de-noising methods that use the wavelet transform, make use of of three steps - In the first step, it computes the wavelet transform of the noisy input image, the second step is used to apply thresholding in order to remove noise on the detailed coefficients and finally inverse wavelet transform is applied of the modified coefficients to get the denoised image. In this paper, the performance of various wavelet shrinkage techniques is reviewed. An Experimental analysis of wavelet based methods including Visu Shrink, Sure Shrink, Bayes Shrink, and a hybrid method (wavelet shrinkage with guided filter) is carried out. For performance comparison, signal to noise ratio, structural similarity index and edge preservation index, parameters are used. From the experimental results, it is concluded that hybrid filter outperform the traditional wavelet shrinkage methods.

Keywords


Speckle Noise, Breast Ultrasound (BUS) image, Wavelet shrinkage, Signal to Noise Ratio (SNR), Structural Similarity Index (SSIM), Edge Preservation Index (EPI).