Multichannel heuristic learning based single layer neural network filter for mixed noise suppression from color Doppler ultrasound images
Manish Kumar, S. Mishra, D. K. Choubey
Jan 22, 2021
Citations
1
Citations
Journal
Journal of Real-Time Image Processing
Full text
Semantic Scholar
Key Takeaway Key Takeaway: The M-Jaya-FLANN neural network effectively suppresses mixed noise from color Doppler ultrasound images, improving patient identification and reducing the impact of speckle noise.
Abstract
Mixed noise suppression from color Doppler ultrasound (CDUS) images is always a challenging task because the noise distribution usually does not have a parametric model and heavy tail. It affects the inherent features of the image awkwardly. Consequently, identifying an internal blockage or hemorrhage of the patient becomes arduous in such conditions. An acquired CDUS image is majorly affected by speckle noise and can be coupled with Gaussian and impulse noises. In this paper, the evolutionary multichannel Jaya based functional link artificial neural network (named as M-Jaya-FLANN) has been proposed to get rid of mixed noise from the CDUS images. The subjective evaluation and the measurement of qualitative metrics, such as structural similarity index, computational time, convergence rate, and Friedman’s test are carried out for the performance analysis of different filters. The research outcomes exhibit the supremacy of the proposed filter over other competitive filters and can handle real-time noise elimination after completion of training. For the experimentation purpose, CDUS image data are collected from Medanta hospital, Ranchi, India.