Content area
Full Text
Abstract- Interpolation is a common technique for predicting unknown data points within the range of discrete known data points. In image processing and analysis applications, image interpolation is employed for changing image resolution but also for tasks like image rotation and other transformations. Classical image interpolation algorithms such as nearest neighbour, bilinear and bicubic interpolation are simple and fast. However, due to the high distortion along image details such as edges, the resulting images are often low in quality. In order to reduce these distortions and preserve fine image details, we propose an edge preserving interpolation algorithm in this paper. Our proposed algorithm extracts and recognises the direction of edges in an image. Based on the extracted information about localisation of edges, interpolated pixels are either replicated or predicted from known neighbourhood pixels. Experimental results confirm our approach to give good image quality, outperforming various other interpolation algorithms.
Keywords: Image interpolation, edge preservation, image detail, upsampling.
(ProQuest: ... denotes formulae omitted.)
1. Introduction
Image interpolation is a technique often employed for obtaining a high resolution (HR) image from a low resolution (LR) image. For such an upsampling process, image interpolation can also be interpreted as increasing the pixel density per unit area of an image to obtain a HR version from a LR counterpart. As illustrated in Fig. 1, if an image is assumed to be a sequence f (xK ) of length N and this sequence is downsampled by a factor of 2 to obtain another sequence g (xn ) of length N/2, then interpolation will yield a sequence l(xk) that should approximate f (xk).
Among the most popular image interpolation algorithms are nearest neighbour interpolation [1], bilinear interpolation [2], bicubic interpolation [3], and spline interpolation [4]. In the nearest neighbour method, the value of a new pixel simply takes on the value of the nearest original pixel. In bilinear interpolation, the interpolated value is calculated as the weighted average of its neighbours. While this takes only four pixels into account, bicubic interpolation considers 16 pixels and cubic fitting functions, whereas spline interpolation is based on a special type of piecewise polynomial interpolant. However, these algorithms, despite being fast and simple, suffer from resulting in relatively low image quality due to aliasing...