Wavelet-based marker controlled watershed transformation

A new method for the generation of superpixels which can be implemented using watershed transformation and also threshold based estimation for image denoising in the wavelet transform domain. Recovery of the image from its noisy atmosphere is also an area of interest in this paper. Our method aims at the extraction process of local and global impression of a given image by giving priority to image denoising also. We propose a gradient-based segmenting adherity property of the segmented image. We also show this as an efficient method to achieve the regularity and adherence property of the segmented image. Since we are dealing with marker controlled Watershed transformation technique, the problem of over-segmentation can be avoided to a great extreme. Partial-thresholding to smoothness and preservation of better image details can be also seen in this work. Better PSNR, MSNR and correlation parameter can be achieved through wavelet thresholding. Here, we try to showcase `markerpixels' as an efficient tool for the creation of superpixels using Watershed transformation and also wavelet transformation process for denoised waterpixel segments. Thus a combination of wavelet based-controlled watershed segmentation can be seen through this work. Soft-thresholding and hard-thresholding are used for image denoising. By this work, it is shown that this method offer better PSNR and MSE.