A FAST AND ROBUST FRAMEWORK FOR IMAGE FUSION AND ...

x
x
1
3
x
x
2
4
y1 y2
y y
3
4
a b c d e
Figure 1.2: An illustrative example of the motion-based super-resolution problem. (a) A high-resolution
image consisting of four pixels. (b)-(e) Low-resolution images consisting of only one pixel, each captured
by subpixel motion of an imaginary camera. Assuming that the camera point spread function is
known, and the graylevel of all bordering pixels is zero, the pixel values of the high-resolution image can
be precisely estimated from the low-resolution images.
use of image processing algorithms, which presumably are relatively inexpensive to implement.
The basic idea behind super-resolution is the fusion of a sequence of low-resolution
(LR) noisy blurred images to produce a higher resolution image. The resulting high-resolution
(HR) image (or sequence) has more high-frequency content and less noise and blur effects than
any of the low-resolution input images. Early works on super-resolution showed that it is the
aliasing effects in the low-resolution images that enable the recovery of the high-resolution
fused image, provided that a relative sub-pixel motion exists between the under-sampled input
images [8].
The very simplified super-resolution experiment of Figure 1.2 illustrates the basics of
the motion-based super-resolution algorithms. A scene consisting of four high-resolution pixels
is shown in Figure 1.2(a). An imaginary camera with controlled subpixel motion, consisting
of only one pixel captures multiple images from this scene. Figures 1.2(b)-(e) illustrate these
captured images. Of course none of these low-resolution images can capture the details of the
underlying image. Assuming that the point spread function (PSF) of the imaginary camera is a
known linear function, and the graylevel of all bordering pixels is zero, the following equations
relate the the low-resolution blurry images to the high-resolution crisper one.
3