Learning-Based Single Image Super-Resolution with Improved Edge Information
- PDF / 2,871,195 Bytes
- 10 Pages / 612 x 792 pts (letter) Page_size
- 27 Downloads / 217 Views
Learning-Based Single Image Super-Resolution with Improved Edge Information G. Mandala,* and D. Bhattacharjeeb,** a
Department of Computer Science and Engineering, Bankura Unnayani Institute of Engineering, Bankura, 722146 India b Department of Computer Science and Engineering, Jadavpur University, Kolkata, 700032 India * e-mail: [email protected] ** e-mail: [email protected] Abstract—A new learning-based single image super-resolution technique that upscales the low resolution (LR) image in a single pass toits desired high resolution (HR) image is proposed here.Inthe upscaling procedure, a linearmapping function is learned from the external data set. Mapping function converts LR patch to its corresponding patch.In most of the patch-based learning technique, smoothness of the overlapped regions is performed with an average value of the overlapped regions. As a result, edge information that reflects in adjacent LR patches does not always transparently reflects in HR patches. So in our technique, we applied edge directed smoothness in adjacent patches. An edge exists along the direction, where the second-order derivative is lower. To reach this, we have selected non-overlapping patch, and after getting HR patch, we performed edge directed smoothness of adjacent patches. This results smoothness of adjacent patches with more detailedge information. Apart from this nonoverlapping patch selection reduces computational complexity, without compromising image quality. Experimental results show significant improvement in terms of subjective and objective quality than other popular learning or interpolation based method.Our method showsrobustness on noisy images also. Keywords: external learning, super-resolution, direct mapping, iterative-curvature, structured-similarity DOI: 10.1134/S1054661820030189
1. INTRODUCTION Image superresolution is a technique that increases the resolution of a low resolution image. Nowadays, a vast number of electronic imaging applications as pattern recognition, medical diagnosis, video surveillance, remote sensing, biometric identification, etc. require a high resolution image to process. However, high resolution images are not always easy to obtain due to the limitations of the image acquisition system or surrounding environmental condition. So, the image superresolution technique becomes a popular topic in modern image processing application. Superresolution method primarily categorized into two techniques, according to the availability of the number of input images it can be divided into two types, namely, single image super-resolution and multi image superresolution techniques [1–3]. In the multi image technique, more than one LR input image is available, and multiple images are the different views of the same image. Single image superresolution method is considered as an ill posed problem as there can be many high resolution images corresponding to a given single low resolution image. Here, we have
Received February 22, 2019; revised December 16, 2019; accepted March 6, 2020
Data Loading...
