Directionalwavelet transform for image denoising

This paper introduces a technique for image denoising based on the one-dimensional wavelet transform computed along several directions on the image. Denoising is implemented using either adaptive or non-adaptive thresholding of the wavelet coefficients. This directional wavelet transform technique was inspired on ridgelet and curvelet transforms. We explore redundancy of the wavelet transform and its property to easily detect singularities to remove noise without smearing the edges in the image. Denoising is improved at increased computational cost. Our denoising technique provides better results than methods like undecimated two dimensional wavelet transform and curvelet transforms, and comparable results to wavelet-based hidden Markov tree method.