Image Zooming Using Corner Matching
This work was intended to direct the choice of an image interpolation/zoom algorithm for use in UND's Open Prototype for Educational Nanosats (OPEN) satellite program. Whether intended for a space-borne platform or a balloon-borne platform, we expect to use a low cost camera (Raspberry Pi) and expect to have very limited bandwidth for image transmission. However, the technique developed could be used for any imaging application. The approach developed analyzes overlapping $3\times 3$ blocks of pixels looking for “L” patterns that suggest the center pixel should be changed such that a triangle pattern results. We compare this approach against different types of single-frame image interpolation algorithms, such as zero-order-hold (ZOH), bilinear, bicubic, and the directional cubic convolution interpolation (DCCI) approach. We use the peak signal-to-noise ratio (PSNR) and mean squared error (MSE) as the primary means of comparison. In all but one of the test cases the proposed method resulted in a lower MSE and higher PSNR than the other methods. Meaning this method results in a more accurate image after zooming than the other methods.