Image Super-resolution Enhancement Research Articles

Image Quality Assessment: Measuring Perceptual Degradation via Distribution Measures in Deep Feature Spaces.

This study aims to develop advanced and training-free full-reference image quality assessment (FR-IQA) models based on deep neural networks. Specifically, we investigate measures that allow us to perceptually compare deep network features and reveal their underlying factors. We find that distribution measures enjoy advanced perceptual awareness and test the Wasserstein distance (WSD), Jensen-Shannon divergence (JSD), and symmetric Kullback-Leibler divergence (SKLD) measures when comparing deep features acquired from various pretrained deep networks, including the Visual Geometry Group (VGG) network, SqueezeNet, MobileNet, and EfficientNet. The proposed FR-IQA models exhibit superior alignment with subjective human evaluations across diverse image quality assessment (IQA) datasets without training, demonstrating the advanced perceptual relevance of distribution measures when comparing deep network features. Additionally, we explore the applicability of deep distribution measures in image super-resolution enhancement tasks, highlighting their potential for guiding perceptual enhancements. The code is available on website. (https://github.com/Buka-Xing/Deep-network-based-distribution-measures-for-full-reference-image-quality-assessment).

Single Image Super Resolution Enhancement Method Based on Fractal Interpolation of Iteration Function System

Based on the similarity and transitivity of local features of image data, self similarity and scale invariance of fractal features, fractal method of Iterated Function System is used to construct fractal interpolation function, and variable vertical scale factor is used to realize super-resolution enhancement of two-dimensional image data. The experimental results show that fractal interpolation method of iterated function system can keep the image details better, and the reconstruction error is smaller, which is better than bilinear interpolation and bicubic interpolation methods.

A nonlinear fourth-order PDE for multi-frame image super-resolution enhancement

The multiframe super-resolution (SR) techniques are considered as one of the active research fields. More precisely, the construction of the desired high resolution (HR) image with less artifacts in the SR models, which are always ill-posed problems, requires a great care. In this paper, we propose a new fourth-order equation based on a diffusive tensor that takes the benefit from the diffusion model of Perona-Malik in the flat regions and the Weickert model near boundaries with a high diffusion order. As a result, the proposed SR approach can efficiently preserve image features such as corner and texture much better with less blur near edges. The existence and uniqueness of the proposed partial differential equation (PDE) are also demonstrated in an appropriate functional space. Finally, the given experimental results show the effectiveness of the proposed PDE compared to some competitive methods in both visually and quantitatively.

Landsat Super-Resolution Enhancement Using Convolution Neural Networks and Sentinel-2 for Training

Landsat is a fundamental data source for understanding historical change and its effect on environmental processes. In this research we test shallow and deep convolution neural networks (CNNs) for Landsat image super-resolution enhancement, trained using Sentinel-2, in three study sites representing boreal forest, tundra, and cropland/woodland environments. The analysis sought to assess baseline performance and determine the capacity for spatial and temporal extension of the trained CNNs. This is not a data fusion approach and a high-resolution image is only needed to train the CNN. Results show improvement with the deeper network generally achieving better results. For spatial and temporal extension, the deep CNN performed the same or better than the shallow CNN, but at greater computational cost. Results for temporal extension were influenced by change potentiality reducing the performance difference between the shallow and deep CNN. Visual examination revealed sharper images regarding land cover boundaries, linear features, and within-cover textures. The results suggest that spatial enhancement of the Landsat archive is feasible, with optimal performance where CNNs can be trained and applied within the same spatial domain. Future research will assess the enhancement on time series and associated land cover applications.

Image Edge Orientation Estimation via Fuzzy Logic

This paper proposes a novel image edge orientation estimation using fuzzy logic approach. The idea behind orientation estimation is to design edge directed image interpolation for image super resolution enhancement. The calculation of image edge pixel orientation is based on different types of gradients measured across different angles from particular edge pixel. Orientation of edge pixel is classified in to nine angles to ensure the quality of orientation estimation. The advantage of proposed fuzzy based method is that, orientation calculation is done in human language with the help of simple fuzzy rules. Use of simple membership function and rules helps to design an efficient orientation estimation frame work. The experimental results show that fuzzy based orientation estimation outperforms even if the image is affected by noise.

基于多核DSP的图像增强实现

针对移动便捷安检设备快速识别物体的需求，我们设计实现了一种基于TMS320C6678的图像超分辨率增强系统。该系统首先采用双三次插值算法放大原始图像，然后采用暗通道算法进行图像增强，并对暗通道算法中的高斯滤波处理部分进行了OpenMp多核加速处理。实现结果表明，增强后的图片特征清晰，便于观察；且利用多核DSP增强后的时间缩短，相当于单核的0.49倍。 To meet the demand of mobile security inspection equipment to rapidly recognize objects, an image super-resolution enhancement system based on TMS320C6678 is designed and realized in this paper. This system first uses bi-cubic interpolation to magnify the image, then uses dark channel algorithm to enhance the image, and adopts OpenMp multi-core acceleration processing to accelerate the Gauss filter processing part of the dark channel algorithm. The results show that the enhanced images are clear to be observed. And the processing time is shortened, which is equivalent to nearly half of the single core.

Image Super Resolution Enhancement Based on Interpolation of Discrete and Stationary Wavelet Domain

In this paper, we propose an image super resolution enhancement technique based on interpolation of the high frequency sub band images obtained by discrete wavelet transform (DWT) using different types of wavelets such as Daubechies 1. Daubechies 2 .., Daubechies 9 haar, and the input image. The edges are enhanced by introducing an intermediate stage by using stationary wavelet transform (SWT). We compare the results of different types of wavelets. DWT is applied in order to decompose an input image into different sub bands. Then the high frequency sub bands as well as the input image are interpolated. The estimated high frequency sub bands are being modified by using high frequency sub bands obtained through SWT then all these sub bands are combined to generate a new super-resolved image by using inverse DWT.