Resolution enhancement of images using wavelet transform extrema extrapolation

Abstract

One problem of image interpolation refers to magnifying a small image without loss in image clarity. We propose a wavelet based method which estimates the higher resolution information needed to sharpen the image. This method extrapolates the wavelet transform of the higher resolution based on the evolution of the wavelet transform extrema across the scales. By identifying three constraints that the higher resolution information needs to obey, we enhance the reconstructed image through alternating projections onto the sets defined by these constraints.

An Introduction to Frames

“…Both Laine and Unser used frames to decompose textures in order to characterize them across scales [124,162]. In [47] the authors use frames for image interpolation and resolution enhancement. In [48], the authors use frames to significantly improve the classification accuracy of protein subcellular location images, to close to 96% as well as the high-throughput tagging of Drosophila embryo developmental stages [112].…”

Adaptive Cycle Spinning Cellular Neural Network for Image Resolution Enhancement

“…It has been widely used in many image-processing applications such as super resolution, 4K TV, high-quality printing, medical imaging, and so on. Recently, image resolution enhancement methods in the wavelet domain have been discussed in many papers [1]- [3]. In wavelet based techniques, it is assumed that the low-resolution (LR) image whose resolution is to be enhanced is a lowpass-filtered and decimated highresolution (HR) image.…”

“…In other words, the LR image corresponds to the low-frequency coefficients of a wavelet-transformed HR image. These methods are based on wavelet-domain zero padding (WZP) [1]. The WZP sets the LR image to a low-frequency subband of a wavelet transform and high-frequency subbands are composed of all-zero matrices, then the interpolated image is reconstructed by an inverse wavelet transform.…”

Image interpolation using wavelet-based contour estimation

“…These are typically ad hoc methods and the outcome is significantly affected by the mistakes made in edge locations. Wavelet [6] or fractal [7] based techniques enhance the high-frequency content of the interpolated image by predicting the high-frequency details in the wavelet or fractal domain. Since no explicit edge model is involved, these techniques usually end up increasing any noise present in the unmagnified image.…”