Abstarct: Super-resolution is a promising technique of digital imaging which attempts to generate a raster image with a higher resolution than its source. The source may consist of one or more images of a scene. This thesis focuses on single-frxame super-resolution i.e. the source is a single raster image. In interpolation methods increasing resolution is performed by upsampling the image. In these methods missing details are not reconstructed properly. In comparison to other image enhancement techniques, super-resolution image reconstruction technique not only improves the quality of under-sampled, low-resolution images by increasing their spatial resolution but also attempts to filter out distortions. Increasing resolution in image means increasing the number of pixels in it. To increase the resolution of the input image we propose a new approach baxsed on self learning technique. In proposed method for learning relation between low/high resolution patches, we use image pyramid that can estimate high frequncy details of the image without need for external databaxse. Using the proposed method in this thesis, user can overcome to problems imaging systems using moderate cost hardware. In other hand user can generate high resolution image without collecting external databaxse that includes high and low resolution images. In this thesis different methods of super resolution will be explained. The proposed method is fast so that for generating high resolution image the average time is 5 second. Also in terms of PSNR a satisfactory result is obtained. Experimental results show that the proposed method can remove artificial effects and also produce sharp edges. The proposed method in terms of time, PSNR and SSIM criteria is better than the other methods intruduced in this thesis