Comparative Study of Feature Extraction Using Different Transform Techniques in Frequency Domain

The compressed sensing is a mathematical approach of reconstructing a signal that is acquired from the dimensionally reduced data coefficients/less number of samples, i.e., less than the Niquist rate. The data coefficients are high-frequency components and low-frequency components. The high-frequency components are due to the rapid changes in the images (edges) and low-frequency correspond are due to slow varying information (continuous surface). The idea is to retain only low-frequency components, i.e., the significant components that constitute the compressed signal. This compressed signal is the sparse signal which is so helpful during medical scenario. During the Medical Resonance Imaging (MRI) scans, the patient undergoes many kinds difficulties like uncomfortness, patients are afraid of the scanning devices, he/she cannot be stable or changing his body positions slightly. Due to all these reasons, there can be a chance of acquiring only the less number of samples during the process of MRI scan. Even though the numbers of samples is less than the Nyquist rate, the reconstruction is possible by using the compressed sensing technique. The work has been carried out in the frequency domain to achieve the sparsity. The comparative study is done on percentage of different levels of sparsity of the signal. This can be verified by using peak signal-to-noise ratio (PSNR), root mean square error (RMSE), and structural similarity (SSIM) methods which are calculated between the reference image and the reconstructed image.