| In this thesis we describe a noise detection and a motion-noise separation algorithm, as well as the stochastic properties of the noise. The difference between corresponding pixels subject to one type of noise, of two frames, has mean vector equal to (0,0,0), and variance covariance matrix with relatively small variances, for the (R, G, B) difference values. The other type of noise is a result of disturbance of the light equilibrium due to motion in neighboring or nearby pixels. In this type of noise the mean of the difference is non-zero. Every pixel not included in the one type of noise or the other is part of the motion set between the two frames. The pixels are organized in macroblocks, so macroblocks containing pixels with motion are applied motion estimation and motion compensation methods first and subsequently the difference between the corresponding macroblocks of the two frames is obtained.; This thesis furthermore describes an algorithm of cancer recognition of ultrasound images. (Abstract shortened by UMI.)... |
