NOISE REMOVAL AND CLASSIFICATION OF ULTRASOUND IMAGE USING MACHINE LEARNING ALGORITHM

Abstract

A clamor decreasing technique dependent on shearlet change and a Laplacian pyramid
channel is displayed. A non subsampled laplacian pyramid channel is utilized to break down the loud
picture in this way deteriorating the picture into high-recurrence and low-recurrence subbands. In
view of the limit work and the relationship of the shearlet coefficients in the transformation domain,
an improved edge shrinkage calculation is proposed to play out the edge shrinkage preparing on the
shearlet coefficients of the high-recurrence subbands. The low-recurrence subbands in the change
space are prepared by the Laplacian pyramid channel, and a denoised ultrasonic image is acquired
by the inverse transformation of the shearlet. At that point the spatial regularization on superpixels to
make divided areas increasingly reduced. The division pipeline involves the calculation of superpixels
and the calculated pixels are used for the extraction of descriptors, for example, shading and surface
and it is followed by delicate grouping, utilizing a standard classifier for regulated learning and the
image is segmented using graph cut methodology. We utilize this division pipeline on four real-time
applications in therapeutic imaging. The proposed strategy was applied to 31 genuine PCOS
ultrasound pictures got from patients and contrasted and three techniques.