Image De-noising Using Intelligent Parameter Adjustment

Abstract

Image de-noising is one of the main steps in the medical image analysis process. In medical imaging, noise usually occurs at the capture stage of medical machines such as the ultrasound machines. This noise may hide important information that affects the diagnosing process. Current medical image de-noising techniques still need modifications to enhance their de-noising capabilities, especially traditional parameter dependent techniques such as VisuShrink. This technique has a threshold that needs to be adjusted to efficiently de-noise the images. In this paper, an intelligent framework is proposed to assign a threshold to VisuShrink technique based on the current image features. These features extracted from the image using Scale Invariant-Feature Transform (SIFT) technique are used to train different machine learning (ML) techniques for predicting the appropriate threshold. The experimental results showed that the proposed framework managed to reduce the noise compared to VisuShrink technique with a fixed threshold.

Binary Descriptors for Dense Stereo Matching

“…RGB images may not be available to be fed to the deep neural network [23]. Furthermore, engineered descriptors can be used with machine and deep learning [24], [25].…”