April 1995, Volume 4, Number 1
This paper presents a new idea of a two layer hybrid image coder based on visual pattern image coding of original image and residual image by means of the wavelet transform. At first contours are extracted from the original image. As a contour extractor a VP1C coder is used. In the second, the residual image is computed and coded by the wavelet transform. At the decoder side the sum of contours and residual (texture) image parts is made to obtain the reconstructed image. Some results of coding simulation on picture Lena (256x256x8) are presented. The proposed coding technique is well suitable for image coding and progressive image transmission.
In this paper a new neural ADC design is presented, which is based on the idea to replace all functional components needed in the ADC block scheme by a simple connection of neurons. Transformation of ADC functional scheme into an analog neural structure and its computer simulation is one of the main results of this paper. Furthermore, a discrete component prototype of the proposed A/D converter is discussed and experimental results are also given.
This paper deals with a large class of nonlinear digital filters, the stack filters, which contain all combinations and compositions of rank order operators within a finite window. Attention is given to design and effective hardware implementation of an optimal stack filter for image processing. Presented simulation results confirm robustness of stack filters in the image restoration corrupted by impulsive noise.
This paper presents the use of B spline functions in various digital signal processing applications. The theory of one-dimensional B spline interpolation is briefly reviewed, followed by its extending to two dimensions. After presenting of one and two dimensional spline interpolation, the algorithms of image interpolation and resolution increasing were proposed. Finally, experimental results of computer simulations are presented.
Two new fast tracking exponentially weighted conventional recursive least-squares (RLS) algorithms of adaptation of the adaptive Volterra filters (AVF) for time-varying systems are presented. The new algorithms are based on a modification of a priciple of variable forgetting factor with unity zone . Their additional computational complexity due to the forgetting factor adaptation is negligible compared to that of conventional RLS algorithms. The performance properties of the proposed algorithms are verified via computer experiments.