Parallel image transmission and compression Using windowed kriging interpolation

Abstract

We propose the use of windowed kriging to enable ordinary kriging to interpolate sampled natural images of much higher resolution. Due to its high computational cost, applying

FPGA Architecture for Kriging Image Interpolation

“…The step for the windowed Kriging interpolation [11] is not very difficult since all the input points to be considered are included in the well-established frame of interest of D×D pixels. The optimal size of D is a subject of research and will be provided later in this paper.…”

“…2 ] where z(s i ) are observations in two different locations separated by h. A variogram is obtained from an experimental semivariance as a result of its fitting to one of variogram models [7], [11]. Experimental semivariance for each distance h in an interpolated image frame may be obtained as…”

“…where λ i and w(x i ) are the coefficient vectors and the values at the known points, respectively [11]. The equation for computing λ i in matrix formulation is given by…”

“…Unfortunately only a part of all the FPGA resources can be devoted to KIM unless there is a separate FPGA dedicated only for its implementation. Fig. 14 presents a block diagram of an exemplary KIM which is capable of processing three windows in parallel [11]. The module works as follows.…”

“…14 is scalable, an arbitrary number of processing streams may be chosen. The optimal performance is achieved when the number of the processing streams is equal to the number of windows [11].…”