Abstract
In this paper, we propose a method of image resolution enhancement using a single image. The proposed method enhances the resolution by emphasizing the high frequency of a magnified low resolution image. Experimental results show that the performance of our method has been improved by 13%-25% over the conventional interpolation methods.
I. INTRODUCTION
Recently, many users require high resolution images as applications and demands for high quality images are increasing. Moreover, we often transfer images by using mobile devices because of improving the performance of hardware. However, when the captured images with high resolution are transferred through the mobile devices, the images are compressed or scaled down in order to reduce the amount of data transmission. In this case, though there are no problems for displaying the transferred images on other devices, the images are actually damaged. Therefore, the technique for enhancing the image resolution is necessary. Image interpolation and super-resolution image reconstruction are generally used as a method for magnifying images. The conventional image interpolation methods such as nearest neighbor, bilinear and bicubic interpolation are suitable for real-time processing because of its speed. However, since image interpolation methods use one low resolution image, it is hard to estimate the high resolution image and there are some limitations caused by missing high frequency components when a low resolution image is generated. Therefore, when an image is simply magnified, the quality of the image will be degraded without information about the original image. In addition, the conventional superresolution methods have complex iterative calculation and many sample images are required for the image reconstruction. For that reasons, it is necessary to develop a technique of image magnification that is simple and uses a few sample images. In this paper, we propose a novel technique for reconstructing a high resolution image from a single low resolution image by emphasizing the high frequency of the input image. II. THE PROPOSED ALGORITHM Fig. 1 shows the overall flowchart of the proposed method. This work was supported by National Research Foundation of Korea Grant funded by the Korean Government (No.2009-0077434).