FAST EDGE-GUIDED INTERPOLATION OF COLOR IMAGES
Amin Behnad and Konstantinos N. Plataniotis
The Edward S. Rogers Sr. Department of Electrical and Computer Engineering, University of Toronto
10 King’s College Road, Toronto, Ontario, M5S 3G4, Canada
Keywords:
Color images, Interpolation, Real-time applications.
Abstract:
We propose a fast adaptive image interpolation method for still color images which is suitable for real-time
applications. The proposed interpolation scheme combines the speed of fast linear image interpolators with
advantages of an edge-guided interpolator. A fast and high performance image interpolation technique is
proposed to interpolate the luminance channel of low-resolution color images. Since the human visual system
is less sensitive to the chrominance channels than the luminance channel, we interpolate the former with the
fast method of bicubic interpolation. This hybrid technique achieves high PSNR and superior visual quality
by preserving edge structures well while maintaining a low computational complexity. As verified by the
simulation results, interpolation artifacts (e.g. blurring, ringing and jaggies) plaguing linear interpolators are
noticeably reduced with our method.
1 INTRODUCTION
Image interpolation has been an active research topic
since early days of image processing, due to a wide
range of its applications, including resolution upcon-
version, resizing, video deinterlacing, video frame
rate upconversion, subpixel motion estimation, im-
age compression, etc. Many of these applications
have real-time requirements, for examples, video de-
interlacing and resolution or/and frame rate upcon-
version. The common solutions for real-time image
interpolation are simple image-independent linear fil-
ters, such as bilinear interpolator and bicubic inter-
polator (Keys, 1981). But these simple linear filters
are isotropic and ill suited for directional image wave-
forms and also cannot cope with the nonlinearities of
the image formation model (Plataniotis et al., 1999).
Hence they tend to produce severe interpolation arti-
facts in areas of edges and fine image details.
To overcome the above said weaknesses of signal-
independent isotropic interpolators, adaptive nonlin-
ear interpolators were introduced (Li and Orchard,
2001), (Muresan, 2005), (Zhang and Wu, 2006) and
(Li and Nguyen, 2008) . Among these algorithms,
edge preserving interpolators are of great interest.
The ultimate goal of an edge-guided image inter-
polation technique is to avoid interpolation against
the existing edge directions for each missing high-
resolution (HR) pixel. This achieves clean and sharp
reproduced edges in the HR output image. How-
ever in practice there is a major issue with most of
the developed edge-guided interpolators. The edge-
guided interpolators achieve better perceptual quality
than linear filters at cost of higher computationalcom-
plexity. Therefore most of these algorithms are not
suitable for real-time applications.
In this work, we address this issue and devise
an algorithm to reduce the computational cost of di-
rectional image interpolation for color images. In
the proposed algorithm, low-resolution (LR) color
images are converted to the luminance-chrominance
space, from the RGB counterpart and the interpola-
tion process is carried out in the new space. The rea-
son for this mapping is twofold. First, the human vi-
sual system is much more sensitive to the luminance
component than the chrominance. Hence, we apply
a sophisticated interpolation technique to interpolate
the luminance channel and for computational savings,
a simple linear filter e.g. bicubic is applied for the
chrominance channels. Second, the luminance chan-
nel captures the variations in the image and magni-
fies the edges and other high-frequency components.
This achieves more reliability on the extracted edge
information from the LR image, which is crucial for
directional image interpolation.
To interpolate the luminance channel, we also pro-
pose a fusion-based image interpolation method. As
verified by the simulations, this technique achieves
superior performance than the competing methods, at
lower computational complexity.
103
Behnad A. and N. Plataniotis K. (2010).
FAST EDGE-GUIDED INTERPOLATION OF COLOR IMAGES.
In Proceedings of the International Conference on Signal Processing and Multimedia Applications, pages 103-107
DOI: 10.5220/0002933401030107
Copyright
c
SciTePress