24-07-2012, 12:32 PM
A New Single Image Interpolation Technique For
Super Resolution
A New Single Image Interpolation Technique.pdf (Size: 2.8 MB / Downloads: 28)
INTRODUCTION
Lens employed in an inexpensive camera usually has
limited spatial resolution. Therefore, for obtaining high
resolution images, an expensive optical imaging system is
employed. However, limitation of space and high cost of
optics is a major hindrance for wide applicability of high
resolution devices. Limitations of optical imaging systems can
be improved by the use of signal processing algorithms. In
signal processing, the quality of an image can be improved by
using approaches such as image enhancement, de-blurring and
super resolution. Interpolation is a technique in which the
function for calculating unknown data points between known
data points is determined. Visual quality of an image is highly
dependent on interpolation technique. The interpolation
techniques are further divided in to two broad categories;
deterministic and statistical interpolation.
EXPERIMENTAL SETUP AND RESULTS
The proposed Interpolation algorithm has been tested on
Intel Core Duo Processor, 2.0 GHz CPU with 1 GB RAM
using MATLAB R2008a version. A threshold value is set
using an iterative process that differentiated between edges
and smooth areas. Edge interpolation is processed covariance
based interpolation, while smooth areas are interpolated
through curvature based approach.
CONCLUSION
In this paper a new algorithm for super resolution of single
image has been presented. This new algorithm is a hybrid
approach that performs interpolation while switching between
covariance based interpolation and iterative curvature based
interpolation. The new approach depicts reduced processing
time and better visual perceptibility. The performance of the
proposed approach was compared with another hybrid
approach iNEDI algorithm and other interpolation techniques
like nearest neighbor, bilinear and bicubic interpolation
techniques. The proposed method showed significantly
improved performance in terms of processing time, peak
signal to noise ratio and visual quality.