Abstract
We present a novel hybrid scheme called "hyper-resolution" that integrates image probability-filtering-based interpolation and improved Poisson maximum a posteriori (MAP) super-resolution to respectively enhance high spatial and spatial-frequency resolutions of a single down-sampled image. A new approach to interpolation is proposed for simultaneous image interpolation and smoothing by exploiting the probability filter coupled with a pyramidal decomposition and the Poisson MAP super-resolution is improved with the techniques of edge maps and pseudo-blurring. Simulation results demonstrate that this hyper-resolution scheme substantially improves the quality of a single gray-level, color, or noisy image, respectively.
Highlights
The quality of an image can be evaluated on its spatial and spatial-frequency resolutions
With the comparison between B-spline and probability-filtering-based interpolation, it can be found that bilinear or cubic spline interpolation performs on an image as weighted mean filtering while the probability-filteringbased interpolation acts as an adaptive filtering useful for preserving image features in addition to image interpolation; that is because the former creates interpolated pixels by averaging their own neighborhood pixels and the latter analyzes the characteristics of surrounding pixels prior to interpolating pixels
The interpolation was done with a factor of 22 for each cycle; the super-resolution procedure of the proposed hyper-resolution scheme, shown in Figure 3, adopted 25 iterations and the point spread function (PSF) with a standard deviation of 1 pixel because of the pyramidal decomposition used here, four neighboring pixels in the interpolated image mainly contributing to one pixel in its own decimated image
Summary
The quality of an image can be evaluated on its spatial and spatial-frequency resolutions. Image interpolation is performed by pixel replication in a small neighborhood of each existing pixel This is equivalent to a first-order linear filter. The improved Poisson maximum a posteriori (MAP) super-resolution [23] is performed to reconstruct the high spatial-frequency spectrum of the interpolated image. To illustrate the performance of the proposed scheme, this research has studied the resolution enhancement on gray-level, color, and noisy images, respectively, and comparisons will be made among the superresolution coupled with different interpolators.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
