Image quality improvement for learning-based super-resolution with PCA

Abstract

Previously, we proposed a learning-based superresolution method using the TV regularization method, which significantly reduced image processing time by removing database redundancy. However, there was a problem when noise appeared in reconstructed images because of an excessive reduction in database redundancy.In this paper, we propose a new learning-based superresolution method, where noise is removed by utilizing Principal Components Analysis (PCA). The obtained algorithms significantly reduce the complexity and maintain a comparable image quality. This facilitates the adoption of learning-based super-resolution by motion pictures, e.g., internet and HDTV movies.

A new super resolution method based on combined sparse representations for remote sensing imagery

“…Miura et al 12 proposed a learning based BTV super resolution method and the method used PCA to remove noise. The results indicated that the method can provide a good performance turning NTSC TV signal to HDTV.…”

“…In general, the learning based SR methods work in the following way 12,13 : learning a degeneration model of high resolution images from low resolution images first, then reconstruct the original high resolution images based on the degeneration model. Miura et al 12 proposed a learning based BTV super resolution method and the method used PCA to remove noise.…”

Image Super-Resolution via Wavelet Feature Extraction and Sparse Representation

“…Figure 2 shows the HF sub-bands obtained via DWT for a given image. Principal Component Analysis (PCA) [25] was employed to reduce the data obtained through the sub-bands interpolated to obtain the most relevant features. One training set S l with the same size as the training set S h was obtained.…”