Abstract:

In order to overcome the blood-speckle noise and ambiguous boundary in intravascular ultrasound （IVUS） images, this paper proposes an algorithm for noise reduction and contrast enhancement based on the dyadic wavelet transform. In this algorithm, first, the noise variance is estimated according to the characteristics of IVUS images. Next, a method to calculate the local threshold is proposed based on the decomposed structure of dyadic wavelet transform. Then, the wavelet coefficients in different scales are treated using the shrinkage techniques which combine the soft threshold filtering with the hard one. Finally, a contrast enhancement algorithm is proposed based on the stretching of wavelet coefficient extreme and on the interpolating with Hermite polynomials. Experimental results indicate that, as compared with the existing denoising methods, the proposed algorithm is more practicable because it not only reduces the blood speckle noise but also enhances the image contrast without eliminating image details.

Key words: image processing, blood-speckle noise reduction, contrast enhancement, wavelet transform