| Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系 === 102 === While the demands to the images’ quality are increasing, how to upscaling a low resolution image to a high resolution image without artifacts has become a common issue in the related fields of research. The technique of “image interpolation” is to improve the resolution of a image which is able to preserve the characteristics of natural images.
A lack fidelity image with different degree, such as jagged or blur, is usually caused by a zoom in image without be fill up according to different characteristics. This could be found in traditional interpolation methods, for example Nearest Neighbor Interpolation, Bilinear interpolation, and Bicubic Interpolation. Therefore, many solutions that are based on edged-directed methods had been proposed. Those earlier methods mainly focus on maintaining the detail part of the picture specifically on its edge area by interpolating from the detected points.
The New Edge-Directed Interpolation, somehow offers different views. It assumes that there is geometric duality between an image and its enlarged version, by interpolating it, the visual effect can be appreciated if its image structure is simpler; conversely, an artifact appears when image structure is complex. Afterward, some researcher states that this situation could be modify by using weighted least squares.
In this paper, we provide an interpolation method which is based on a modified weighted least squares. This method can provide an accurate interpolation, which recalculates a new weight by adding an error estimation in the ordinal method. According to the experiment results, it shows much better results than some existing methods.
|
|---|
