Image upscaling through pixel interpolation is used in
many different fields to create high resolution images from low resolution
images that are not riddled with jagged edges or excessive blurriness. Image
resolution limits the extent to which zooming enhances clarity, restricts the
quality of digital photograph enlargements, and, in the context of medical
images, can prevent a correct diagnosis. Interpolation can artificially increase
image resolution but is generally limited in terms of enhancing image clarity or
revealing higher frequency content. Algorithmic designs must balance qualitative
improvements, accuracy, artifacts, and complexity for the desired application.
Edge fidelity is one of the most critical components of subjective image
quality, and a number of edge-preserving and edge directed interpolation methods
have achieved varying degrees of success. However many of the algorithms that
achieve good results, are also computationally expensive and time consuming.
There is a need for an algorithm that can achieve good clarity, as well as be
computationally efficient.
To address these issues, researchers at Arizona State
University introduce a new image resizing algorithm based off of the principles
of optical flow as utilized for inter-frame (video) interpolation. The optical
flow equation insists that for every pixel in a given video frame there exists
an isointense pixel in adjacent frames. For video, this amounts to the assertion
that subsequent frames are reconfigurations of the same pixels. The inventors
apply the optical flow equation to the adjacent rows and columns of single
images. The physical basis for optical flow in video (objects are moving) is
void in the static image application. However, the use of the optical flow
equation in our implementation results in a method superior to the traditional
bilinear and bicubic interpolators and competitive with NEDI and iNEDI and at
much faster speeds and arbitrary scaling factors.
Potential Applications
- Medical imaging
- Image viewing and processing software
- Photographic printing
- Computer graphics
- High fidelity of images
- Faster processing with less computational overhead.
- An 8x expansion of an image took 235 times longer using a
conventional method than the proposed method
- A 2x expansion of an image took 100 times longer using a
conventional method than the proposed method
- Accommodates arbitrary scaling factors