Multi-scale Image Enhancement Research Articles

Adversarial multi-image steganography via texture evaluation and multi-scale image enhancement

Adversarial multi-image steganography via texture evaluation and multi-scale image enhancement

Multi-scale progressive fusion-based depth image completion and enhancement for industrial collaborative robot applications

Multi-scale progressive fusion-based depth image completion and enhancement for industrial collaborative robot applications

A novel coal-gangue recognition method in underground coal mine based on image processing

ABSTRACT This study proposes a coal-gangue recognition method based on Retinex and extreme learning machine. The traditional Retinex has many limitations, including halo artifacts, excessive enhancement, and noise amplification in dark areas. To handle these, first, a modified multi-scale Retinex image enhancement algorithm based on hue, saturation, value color space, and bidimensional empirical mode decomposition named modified multi-scale Retinex algorithm with color restoration (MMSRCR) is proposed herein. On-site images are used to verify the effectiveness of the MMSRCR. Second, a preliminary segmentation method for coal-gangue images based on the optimal entropy threshold optimized by the improved quantum genetic algorithm is designed. The proposed segmentation method performs excellent stability compared with other traditional methods. Third, the adhesion part of the ore is further segmented based on the morphology opening operation and watershed algorithm. Then, the feature space of each enclosed area is extracted using the chain code and line segment tables. The improved extreme learning machine completes the classification of these areas, and the coal-gangue visual ratio recognition method is proposed. Finally, an experimental platform for visual ratio recognition is built to verify the proposed method. The experimental results show that the proposed method can accurately identify the visual ratio of coal-gangue.

An Improved Method for Reconstruction and Enhancing Dark Images based on CLAHE

The obtained images are frequently flawed because of a variety of environmen- tal issues, particularly at night, such as inside illumination, cloudy weather, etc. The dark image has a compressed dynamic range that can be improved in order to see the finer details. This research effort proposes an improved lighting reflection model-based technique for improving extremely dim images. This improved method relies on reconstruction carried out via morphological processing with Top-hat transformation and Contrast Limited Adaptive His- togram Equalization (CLAHE). The HSV colour scheme is used to perceive the image, and the V component is estimated. The inverse of the intensity component (V) is calculated after normalising the intensity component. The negative image is then subjected to the CLAHE algorithm. The final step is to apply multiscale image enhancement to the obtained image. The brightness component of an image is adaptively adjusted via gamma enhancement. The outcomes of two gamma-enhanced photographs with various gamma values are produced after gamma enhancement. The significant information that can be used for image fusion is extracted from these images using principal com- ponent analysis. The weight value is adaptively determined during the PCA- based image fusion employing morphological Top-hat modification to enhance the image quality and highlight the erratic background pixels. The suggested technique emphasises edge and structural preservation while enhancing detail in extremely dim photos. Results from experimental validations demonstrate that the suggested strategy outperforms the current method in terms of both qualitative and quantitative measures.

Halo-free image enhancement through multi-scale detail sharpening and single-scale contrast stretching

Conventional image enhancement methods manipulate images in just one scale. This is usually not enough in practice since details like edges live across scales. So, enhancing images through multiscale manipulations is necessary to achieve satisfying results. One drawback of conventional multiscale enhancement methods is that halo artifacts are easy to be introduced. In this paper, we propose a novel efficient multi-scale halo-free image enhancement method, which is based on the observation that image enhancement can be decomposed into two layers: detail sharpening and contrast stretching, and multiscale operations need to be employed to process just the detail layer. We then propose to decompose the multiscale detail manipulations into one-dimensional edge-preserving operations such that halos are completely avoided. For the contrast stretching layer, a modified histogram equalization method is proposed, which achieves state-of-the-art global contrast enhancing effect. Numerical experiments show that, compared to existing popular approaches, the proposed method produces quite competitive results for a variety of image types.

Low light combining multiscale deep learning networks and image enhancement algorithm

Aiming at the lack of reference images for low-light enhancement tasks and the problems of color distortion, texture loss, blurred details, and difficulty in obtaining ground-truth images in existing algorithms, this paper proposes a multi-scale weighted feature low-light based on Retinex theory and attention mechanism. An image enhancement algorithm is proposed. The algorithm performs multi-scale feature extraction on low-light images through the feature extraction module based on the Unet architecture, generates a high-dimensional multi-scale feature map, and establishes an attention mechanism module to highlight the feature information of different scales that are beneficial to the enhanced image, and obtain a weighted image. High-dimensional feature map, the final reflection estimation module uses Retinex theory to build a network model, and generates the final enhanced image through the high-dimensional feature map. An end-to-end network architecture is designed and a set of self-regular loss functions are used to constrain the network model, which gets rid of the constraints of reference images and realizes unsupervised learning. The final experimental results show that the algorithm in this paper maintains high image details and textures while enhancing the contrast and clarity of the image, has good visual effects, can effectively enhance low-light images, and greatly improves the visual quality. Compared with other enhanced algorithms, the objective indicators PSNR and SSIM have been improved.

Multiscale Low-Light Image Enhancement Network With Illumination Constraint

Images captured under low-light environments typically have poor visibility, affecting many advanced computer vision tasks. In recent years, there have been some low-light image enhancement models based on deep learning, but they have not been able to effectively mine the deep multiscale features in the image, resulting in poor generalization performance and instability of the model. The disadvantages are mainly reflected in the color distortion, color unsaturation and artifacts. Current methods unable to adjust the exposure effectively, resulting in uneven exposure or partial overexposure. To address these issues, we propose an end-to-end low-light image enhancement model, which is called multiscale low-light image enhancement network with illumination constraint (MLLEN-IC), to achieve preferable generalization ability and stable performance. On the one hand, we use the squeeze-and-excitation-Res2Net block (SE-Res2block) as a base unit to enhance the model’s ability by extracting deep multiscale features. On the other hand, to make the model more adaptable in low-light image enhancement tasks, we calculate the illumination constraint by the low-light itself to prevent overexposure, uneven exposure, and unsaturated colors. Extensive experiments are conducted to demonstrate MLLEN-IC not only adjusts light levels, but also has a more natural visual effect, and avoids problems such as color distortion, artifacts, and uneven exposure. In particular, MLLEN-IC has pretty generalization and stability performance. The source code and supplementary are available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/CCECfgd/MLLEN-IC</uri> .

Low-light image enhancement with contrast regularization

Because the processing of existing low-light images undergoes multiple sampling processing, there is serious information degradation, and only clear images are used as positive samples to guide network training, low-light image enhancement processing is still a challenging and unsettled problem. Therefore, a multi-scale contrast learning low-light image enhancement network is proposed. First, the image generates rich features through the input module, and then the features are imported into a multi-scale enhancement network with dense residual blocks, using positive and negative samples to guide the network training, and finally using the refinement module to enrich the image details. Experimental results on the dataset show that this method can reduce noise and artifacts in low-light images, and can improve contrast and brightness, demonstrating its advantages.

Multi-scale underwater image enhancement with optimized homomorphic filter and RGB color correction

Multi-scale underwater image enhancement with optimized homomorphic filter and RGB color correction

Generation of Anatomically Inspired Human Airway Tree Using Electrical Impedance Tomography: A Method to Estimate Regional Lung Filling Characteristics.

The purpose of lung recruitment is to improve and optimize the air exchange flow in the lungs by adjusting the respiratory settings during mechanical ventilation. Electrical impedance tomography (EIT) is a monitoring tool that permits measurement of regional pulmonary filling characteristics or filling index (FI) during ventilation. The conventional EIT system has limitations which compromise the accuracy of the FI. This paper proposes a novel and automated methodology for accurate FI estimation based on EIT images of recruitable regional collapse and hyperdistension during incremental positive end-expiratory pressure. It identifies details of the airway tree (AT) to generate a correction factor to the FIs providing an accurate measurement. Multi-scale image enhancement followed by identification of the AT skeleton with a robust and self-exploratory tracing algorithm is used to automatically estimate the FI. AT tracing was validated using phantom data on a ground-truth lung. Based on generated phantom EIT images, including an established reference, the proposed method results in more accurate FI estimation of 65% in all quadrants compared with the current state-of-the-art. Measured regional filling characteristics were also examined by comparing regional and global impedance variations in clinically recorded data from ten different subjects. Clinical tests on filling characteristics based on extraction of the AT from the resolution enhanced EIT images indicated a more accurate result compared with the standard EIT images.

A Multi-Scale Image Enhancement Algorithm Based on Deep Learning and Illumination Compensation

A Multi-Scale Image Enhancement Algorithm Based on Deep Learning and Illumination Compensation

Spatially adaptive multi-scale image enhancement based on nonsubsampled contourlet transform

Low or uneven luminance results in low contrast of near-infrared and optical remote sensing images, making it challenging to analyze their contents. Traditional image enhancement methods cannot simultaneously take detail preservation, contrast enhancement, and brightness improvement into account. In order to cope with this problem, this paper proposes a spatially adaptive multi-scale image enhancement (SAMSIE) scheme, including three key procedures: First, nonsubsampled contourlet transform (NSCT) is employed to decompose a low-contrast image into multi-scale layers. Second, a spatially adaptive Gamma correction strategy based on improved histogram equalization is proposed to enhance the base layer which is used as a guide layer. Third, an adaptive enhancement operator is proposed to enhance fine details. Finally, a high-contrast optical infrared image is obtained by the inverse NSCT with usage of these enhanced layers. The effectiveness of the proposed SAMSIE method is validated by both visualization assess and the evaluation of three quantitative indexes including discrete entropy (DE), contrast gain (CG), and mean brightness improvement (MBI), with comparison of the state-of-the-arts.

Infrared and visible image fusion based on guided hybrid model and generative adversarial network

To retain the detail information and enhance the visibility of the source images, a new and effective image fusion based on a feature decomposition guided hybrid model and GAN is proposed. Firstly, the source images are divided into low-frequency components and high-frequency components by latent low-rank decomposition. Secondly, a multi-scale image enhancement method is used to improve its detail information for the low-frequency component. The low-frequency components are fused to obtain the base layer image by an improved generative adversarial network method. WLS-based fusion rule is used to improve the contrast of an image and get the detail layer image. Finally, the fused image is reconstructed by the linear combination of final base and detail layer images. Experiments are conducted on public datasets and compared with the state-of-the-art image fusion methods. Our experiments show that the proposed method can achieve good scores based on multiple evaluation indices, such as spatial frequency (SF), peak signal-to-noise ratio (PSNR) and pixel feature mutual information (FMI_pixel).

Segmentation and Automatic Identification of Vasculature in Coronary Angiograms

Coronary angiography is the “gold standard” for the diagnosis of coronary heart disease, of which vessel segmentation and identification technologies are paid much attention to. However, because of the characteristics of coronary angiograms, such as the complex and variable morphology of coronary artery structure and the noise caused by various factors, there are many difficulties in these studies. To conquer these problems, we design a preprocessing scheme including block-matching and 3D filtering, unsharp masking, contrast-limited adaptive histogram equalization, and multiscale image enhancement to improve the quality of the image and enhance the vascular structure. To achieve vessel segmentation, we use the C-V model to extract the vascular contour. Finally, we propose an improved adaptive tracking algorithm to realize automatic identification of the vascular skeleton. According to our experiments, the vascular structures can be successfully highlighted and the background is restrained by the preprocessing scheme, the continuous contour of the vessel is extracted accurately by the C-V model, and it is verified that the proposed tracking method has higher accuracy and stronger robustness compared with the existing adaptive tracking method.

Adaptive Multi-Scale Image Enhancement for Digital Radiography.

Digital radiography has been increasingly adopted since it can provide better image quality compared to conventional screen/film method. However, digital radiography can sometimes produce low-quality images because its processing algorithm is unaware of the content. Here, an adaptive multi-scale image enhancement algorithm for digital radiography is demonstrated. The algorithm adapts to the context of the image, thus providing better image quality. The qualitative and quantitative validations of the algorithm in phantoms and in clinical settings showed satisfactory performance.

Improved Method for Enhancing Dark Images based on CLAHE and Morphological Reconstruction

Abstract The image acquired are often flawed due to various defects from environmental scenarios such as indoor lighting, cloudy weather etc especially at night time. The dark image contains compressed dynamic range that can be enhanced for understanding the detailed information. In this investigative work, an improved method for enhancing extremely dark images is proposed utilizing the concept of illumination reflection model. This improved technique is based on Contrast Limited Adaptive Histogram Equalization (CLAHE) and reconstruction done using morphological processing with Top-hat transformation. The image is perceived in HSV color system and V component is estimated. This intensity (V) component is normalized and the inverse of the intensity component is computed. Then CLAHE algorithm is applied on the negative image. Then multiscale image enhancement is applied to the resultant image. Gamma enhancement is used to adaptively adjust the brightness component of an image. After gamma enhancement, the results of two gamma enhanced images with different gamma values are obtained. Principal Component Analysis extracts the significant information from these images that can be used for image fusion. During the PCA based image fusion, the weight value is adaptively calculated by using morphological Top-hat transformation to improve the quality of the image and illuminate the inconsistent background pixels. The proposed method focuses detail enhancement of extremely dark images by preserving the edges and structures. Experimental validations and results show the better performance of the proposed method than the existing method in terms of qualitative and quantitative measures.

Research on Pedestrian Detection Technology Based on MSR and Faster R-CNN

In order to avoid the problem of poor illumination characteristics and inaccurate positioning accuracy, this paper proposed a pedestrian detection algorithm suitable for low-light environments. The algorithm first applied the multi-scale Retinex image enhancement algorithm to the sample pre-processing of deep learning to improve the image resolution. Then the paper used the faster regional convolutional neural network to train the pedestrian detection model, extracted the pedestrian characteristics, and obtained the bounding boxes through classification and position regression. Finally, the pedestrian detection process was carried out by introducing the Soft-NMS algorithm, and the redundant bounding box was eliminated to obtain the best pedestrian detection position. The experimental results showed that the proposed detection algorithm achieves an average accuracy of 89.74% on the low-light dataset, and the pedestrian detection effect was more significant.

An Improved Multi-Scale Image Enhancement Method Based on Retinex Theory

An Improved Multi-Scale Image Enhancement Method Based on Retinex Theory

基于HSI空间的Retinex低照度图像增强算法

低照度环境下拍摄的图像往往会出现细节模糊不清的情况，传统的Retinex算法可以有效提高图像质量，然而当图像亮度整体较低时可能会出现过增强现象。为了改善低照度环境下拍摄图像的质量，本文提出了一种基于HSI颜色空间的改进的多尺度Retinex图像增强算法，采用引导图像滤波消除应用Retinex算法所产生的“光晕伪影”现象，并利用原图像的亮度信息对多尺度Retinex算法增强后的图像进行了亮度重构，有效地抑制了过增强等失真现象，进一步提高了图像增强质量。实验结果表明，本文算法与传统的多尺度Retinex算法相比在低照度图像亮度增强、对比度增强和图像细节信息保护方面具有更好的表现效果。 The images which are captured in low brightness environment are often blurred. The traditional Retinex algorithm can improve the quality of image; however enhancement phenomenon may occur when the image brightness is low. In order to improve the quality of captured images in brightness environment, an improved multi-scale Retinex image enhancement algorithm based on HSI color space is proposed in this paper. In this algorithm, the guide image filtering is used to eliminate the “halo artifact”. The intensity of enhanced image based on multi-scale Retinex algorithm, is reconstructed by applying the intensity of original image. This effectively suppresses the distortion such as over-enhancement, and further enhances the image quality. Experimental results show that for low-luminance image the proposed algorithm has better performance than the traditional multi-scale Retinex algorithm in brightness enhancement, contrast enhancement and image detail information protection.

Multiscale Image Enhancement Techniques with Shrinkage Denoising based on Characteristics of Human Visual System

Multiscale Image Enhancement Techniques with Shrinkage Denoising based on Characteristics of Human Visual System