Edge Model Based High Resolution Image Generation ⋆ Malay Kumar Nema 1 , Subrata Rakshit 1 , and Subhasis Chaudhuri 2 1 Centre for Artificial Intelligence and Robotics, Bangalore 2 VIP Lab, Department of Electrical Engineering, IIT Bombay, Mumbai Abstract. The present paper proposes a new method for high resolution image generation from a single image. Generation of high resolution (HR) images from lower resolution image(s) is achieved by either reconstruction-based methods or by learning-based methods. Re- construction based methods use multiple images of the same scene to gather the extra information needed for the HR. The learning-based methods rely on the learning of characteristics of a specific image set to inject the extra information for HR generation. The proposed method is a variation of this strategy. It uses a generative model for sharp edges in images as well as descriptive models for edge representation. This prior information is injected using the Symmetric Residue Pyramid scheme. The advantages of this scheme are that it generates sharp edges with no ringing artefacts in the HR and that the models are universal enough to allow usage on wide variety of images without requirement of training and/or adaptation. Results have been generated and compared to actual high resolution images. Index terms: Super-Resolution, edge modelling, Laplacian pyramids. 1 Introduction Generation of high resolution (HR) images from low resolution (LR) images have been attempted through reconstruction based approaches and learning based ap- proaches. Reconstruction based approaches require multiple images. They make use of subpixel shifts between images to pool in the extra information needed to create the HR image. Methods employed include sub-pixel registration, nonuni- form interpolation [1][2] and frequency domain approaches [3][4]. An exhaustive list of methods can be found in [5], [6]. Learning based approaches build a relation between LR and HR images, based on the imaging process and/or description of corresponding edges between LR and HR. Multiresolution based mehods are a natural choice for this problem. The multiresolution representations seperate the information in images by frequency. The generation of HR is essentially the problem of generating the missing (hypothetical) level(-1) subband. Solutions have been proposed based on zoom [7][8], wavelet [9] and contourlet [10] coef- ficients. A detailed discussion can be obtained from [11]. The problem may be ⋆ This work is supported by DRDO through project CAR-008. P. Kalra and S. Peleg (Eds.): ICVGIP 2006, LNCS 4338, pp. 1–12, 2006. c  Springer-Verlag Berlin Heidelberg 2006