Noise removal using double density Complex Dual Tree Transform with NeighShrink SURE and median filter

Image de-noising can be called the process of removing noise from the image naturally contaminated by it. It is well known that during transmission or recieval of image, noise gets added inevitably affecting the quality of image. Hence, to preserve the image quality de-noising operations are performed on it. Today there exist various methods for dealing with this problem, depending on the kind of noise present in the image. Wavelet-based image de-noising techniques possess significant importance in this area. This is due to the fact that wavelets have natural ability to represent images in a very sparse number of coefficients. But they have limited directional selectivity and shift sensitivity. To overcome these shortcomings Dual Tree Complex Transforms were introduced. Also as it is known that over-complete transforms provide better balance between complexity and performance, one of its example is double density wavelets. Hence this paper presents a de-noising method, using double density wavelets for generating complex dual tree for removing Gaussian noise. It thresholds the high sub-band coefficients using NeighShrinkSure thresholding and for further smoothening median filter has been used. Further the work was extended to mixed noise as well. Results show that this method produces better results than many of the existing algorithms for both Gaussian as well as mixed noise. PSNR has been used to compare the results.