FIGURE 16 Theoretical performance index for a first-order Markov model with p = 0,9. Triangles: Interpolating functions. Circles: noninterpolating functions.
To magnify the degradation that results from interpolation, we adopt the following strategy that has for goal the highlighting of—as opposed to the avoidance of—the deleterious effects of interpolation: We apply a succession of r = 15 rotations of 2n = 24° each to some image, such that the output of any given step is used as input for the next step; then, we compare the initial image with the final output. To limit potential boundary effects, we perform the final comparison on the central square of the image only. Also, we avoid any interaction between interpolation and quantization by performing all computations in a floating-point format. Figure 17 shows the image we want to rotate. It is made of a radial sinusoidal chirp with a spatial frequency that decreases from center to edge.
Was this article helpful?
Post a comment