Problem Of Image Enhancement Research Articles

Learning Hadamard-Product-Propagation for Image Dehazing and Beyond

Image dehazing has evolved into an attractive research field in the computer vision community in the past few decades. Previous traditional approaches attempt to design energy-based objective functions. However, they cannot accurately express the intrinsic characteristics of the images, posing weak adaptation ability for real-world complex scenarios. More recently, deep learning techniques for image dehazing have matured and become more reliable, showing outstanding performance. Nevertheless, these methods heavily depend on training data, restricting their application ranges. More importantly, both traditional and deep learning approaches all ignore a common issue, noises/artifacts always appear in the recovery process. To this end, a new Hadamard-Product (HP) model is proposed, which consists of a series of data-driven priors. Based on this model, we derive a Learnable Hadamard-Product-Propagation (LHPP) by cascading a series of principle-inspired guidance and recovery modules. In which, the principle-inspired guidance related to transmission is endowed the smoothness property, the other recovery module satisfies the distribution of natural images. The Hadamard-product-based propagations is generated in our developed learnable framework for the task of image dehazing. In this way, we can eliminate noises/artifacts in the recovery procedure to obtain the ideal outputs. Subsequently, since the generality of our HP model, we successfully extend our LHPP to settle low-light image enhancement and underwater image enhancement problems. A series of analytical experiments are performed to verify our effectiveness. Plenty of performance evaluations on three complex tasks fully reveal our superiority against multiple state-of-the-art methods.

Edge‐aware image filtering using a structure‐guided CNN

Image filtering is a fundamental preprocessing step for accurate, robust computer vision applications such as image segmentation, object classification, and reconstruction. However, many convolutional neural network (CNN)-based methods tend to lose significant edge information in the output layer, and generate undesired artefacts in the feature extraction layers. This study presents a deep CNN model for edge-aware image filtering. The proposed network model consists of three sub-networks: (i) feature extraction, (ii) convolution artefact removal, and (iii) structure extraction networks. The proposed network model has an end-to-end trainable architecture that does not need any post-processing steps. Especially, the structure extraction network can successfully preserve significant edges. The proposed filter outperforms state-of-the-art denoising filters in terms of both objective and subjective measures, and can be used for various image enhancement and restoration problems such as edge-preserving smoothing, image denoising, deblurring, and deblocking.

Mathematical modelling of an airborne technical vision system development

An airborne technical vision system of a modern aircraft is, on the one hand, a complex of multispectral sensors and, on the other, airborne computer equipment. The onboard computer solves a wide range of complex digital image processing tasks. The strict requirements are imposed on the computer mathematical software on the image processing time and on the quality of the result displayed on the pilot screen. The software development for such a complex system involves mathematical modelling and use modern information technologies in process of computational algorithms development for task solving. The proposed work discusses ways of solving subtasks of one of the most important subsystems of the airborne complex - the image fusion subsystem and solving the inverse navigation problem. The subsystem is represented by two blocks of tasks: tasks of primary image processing and high-level tasks. The primary processing unit solves the problems of image noise suppression, image enhancement, image fusion and edge detection. In the high-level block of tasks, the tasks of key points detection on real and virtual images, the tasks of a geometric transformation estimation of one image to the plane of another, and the task of a 3D model construction of the underlying surface are solved. The paper presents the results of mathematical modelling in the process of algorithms development for above problems solving. The new results of mathematical modelling of the noise estimation problem are presented in detail. The description of methods and results of other problems solving is less detailed due to article limit but provided with links to the corresponding publications of the authors.

Robust image retrieval by cascading a deep quality assessment network

The performance of computer vision algorithms can severely degrade in the presence of a variety of distortions. While image enhancement algorithms have evolved to optimize image quality as measured according to human visual perception, their relevance in maximizing the success of computer vision algorithms operating on the enhanced image has been much less investigated. We consider the problem of image enhancement to combat Gaussian noise and low resolution with respect to the specific application of image retrieval from a dataset. We define the notion of image quality as determined by the success of image retrieval and design a deep convolutional neural network (CNN) to predict this quality. This network is then cascaded with a deep CNN designed for image denoising or super resolution, allowing for optimization of the enhancement CNN to maximize retrieval performance. This framework allows us to couple enhancement to the retrieval problem. We also consider the problem of adapting image features for robust retrieval performance in the presence of distortions. We show through experiments on distorted images of the Oxford and Paris buildings datasets that our algorithms yield improved mean average precision when compared to using enhancement methods that are oblivious to the task of image retrieval. 11The material in this paper appeared in part at the 2018 IEEE International Conference on Image Processing, Athens, Greece.

Reference-Guided Deep Super-Resolution via Manifold Localized External Compensation

The rapid development of social network and online multimedia technology makes it possible to address traditional image and video enhancement problems, with the aid of online similar reference data. In this paper, we tackle the problem of super-resolution (SR) in this way, specifically aiming to handle the “one-to-many” problem between the image patches of low resolution (LR) and high resolution (HR). We propose a manifold localized deep external compensation (MALDEC) network to additionally utilize reference images, i. e., retrieved similar images in cloud database and reference HR frame in a video, to provide an accurate localization and mapping to the HR manifold, and compensate the lost high-frequency details. The proposed network employs a three-step recovery: 1) internal structure inference, which uses the LR image itself and the internally inferred high frequency information to preserve main structure of the HR image; 2) manifold localization, which localizes the HR manifold and constructs the correspondence between the internal inferred image and the external images; and 3) external compensation, which introduces the external references of retrieved similar patches based on manifold localization information to reconstruct the high-frequency details. The learnable components of MALDEC, internal structure inference, and external compensation, are trained jointly to make a good tradeoff between these two terms for an optimal SR result. Finally, the proposed method is examined under three tasks: cloud-based image SR, multi-pose face reconstruction, and reference frame-guided video SR. Extensive experiments demonstrate the superiority of our method than the state-of-the-art SR methods in both objective and subjective evaluations, and our method offers new state-of-the-art performance.

Deep learning-based super-resolution in coherent imaging systems

We present a deep learning framework based on a generative adversarial network (GAN) to perform super-resolution in coherent imaging systems. We demonstrate that this framework can enhance the resolution of both pixel size-limited and diffraction-limited coherent imaging systems. The capabilities of this approach are experimentally validated by super-resolving complex-valued images acquired using a lensfree on-chip holographic microscope, the resolution of which was pixel size-limited. Using the same GAN-based approach, we also improved the resolution of a lens-based holographic imaging system that was limited in resolution by the numerical aperture of its objective lens. This deep learning-based super-resolution framework can be broadly applied to enhance the space-bandwidth product of coherent imaging systems using image data and convolutional neural networks, and provides a rapid, non-iterative method for solving inverse image reconstruction or enhancement problems in optics.

Contrast enhancement by multi-level histogram shape segmentation with adaptive detail enhancement for noise suppression

The usual problems associated with image enhancement include over- and under-enhancement, halo effects at edges and the degradation of the signal-to-noise ratio as the enhancement of details increases. Some of those problems manifest in the background and some in the details of the enhanced image. Our proposed method is to apply different techniques to enhance the background and details separately. For enhancement of the image background, a novel multi-level histogram shape segmentation method which will detect abrupt changes in the histogram is proposed so that regions of intensity values with a similar frequency of occurrence are segmented for individual equalization to avoid over-enhancement. For detail enhancement, a novel adaptive median based enhancement method is applied to the details to avoid over- and under-enhancement while suppressing noise by limiting the degree of enhancement in homogeneous regions. Halo effects due to the over-enhancement of edges are avoided in our proposed method by using an edge preserving filter for the separation of the background and details so that edges are excluded from detail enhancement. It has been shown that our proposed method is able to avoid the usual adverse problems of image enhancement while producing adequate overall enhancement.

A Survey on Nature-Inspired Optimization Algorithms and Their Application in Image Enhancement Domain

In the field of image processing, there are several problems where the efficient search has to be performed in complex search domain to find an optimal solution. Image enhancement which improves the quality of an image for visual analysis and/or machine understanding is one of these problems. There is no unique image enhancement technique and it’s measurement criterion which satisfies all the necessity and quantitatively judge the quality of a given image respectively. Thus sometimes proper image enhancement problem becomes hard and takes large computational time. In order to overcome that problem, researchers formulated the image enhancement as optimization problems and solved using Nature-Inspired Optimization Algorithms (NIOAs) which starts a new era in image enhancement field. This study presents an up-to-date review over the application of NIOAs in image enhancement domain. The key issues which are involved in the formulation of NIOAs based image enhancement models are also discussed here.

Image enhancement based on multi-guided filtering

Image enhancement, as a basic image proicessing technique, contains much research content, such as enhance contrast, image restoration, noise reduction, image sharpening, distortion correction, etc. The purpose of image enhancement is to effectively highlight the useful information in target image and suppress noise as well. The conventional image enhancement methods are always powerless to tackle the complicated gradient distributions in natural images, and they are also difficult to retain the information about edges accurately. For improving the status of over-smoothing on boundaries, we propose an image enhancement method based on multi-guided filtering. We first synthetically analyze the property of joint filtering and propose the general image optimization model in which the variable parameter is filter kernel. Different filter kernel in the optimization model above generate different filtering method. That is to say, we can use this model to describe the image enhancement problems. The existing joint filters can be regarded as close form solutions of the optimization model above. Inspired by ensemble theory, we use multiple guided images in joint filtering instead of a single guided image to make full use of structure information. By doing so, the image enhancement based on multi-guided filtering can obtain more accurate filtering results. In order to keep the consistency among the multiple filtering outputs of multi-guided filtering method, we add a regularization term into a general image optimization model. We also take into consideration the consistency of pixels in the same image. The experimental results about the noise reduction and image enhancement show that the image enhancement based on multi-guided filtering can give rise to significant outputs. The peak-signal-to-noise ratio of output image of proposed method is higher than those from the traditional image enhancement methods. Therefore, the image enhancement based on multi-guided filtering can improve the quality of digital images efficiently and effectively. This provides a good precondition for subsequent image processing steps and has a prospect of very wide application.

Learning a No-Reference Quality Assessment Model of Enhanced Images With Big Data.

In this paper, we investigate into the problem of image quality assessment (IQA) and enhancement via machine learning. This issue has long attracted a wide range of attention in computational intelligence and image processing communities, since, for many practical applications, e.g., object detection and recognition, raw images are usually needed to be appropriately enhanced to raise the visual quality (e.g., visibility and contrast). In fact, proper enhancement can noticeably improve the quality of input images, even better than originally captured images, which are generally thought to be of the best quality. In this paper, we present two most important contributions. The first contribution is to develop a new no-reference (NR) IQA model. Given an image, our quality measure first extracts 17 features through analysis of contrast, sharpness, brightness and more, and then yields a measure of visual quality using a regression module, which is learned with big-data training samples that are much bigger than the size of relevant image data sets. The results of experiments on nine data sets validate the superiority and efficiency of our blind metric compared with typical state-of-the-art full-reference, reduced-reference and NA IQA methods. The second contribution is that a robust image enhancement framework is established based on quality optimization. For an input image, by the guidance of the proposed NR-IQA measure, we conduct histogram modification to successively rectify image brightness and contrast to a proper level. Thorough tests demonstrate that our framework can well enhance natural images, low-contrast images, low-light images, and dehazed images. The source code will be released at https://sites.google.com/site/guke198701/publications.

Low lighting image enhancement using local maximum color value prior

We study the problem of low lighting image enhancement. Previous enhancement methods for images under low lighting conditions usually fail to consider the factor of image degradation during image formation. As a result, the lost contrast could not be recovered after enhancement. This paper will adaptively recover the contrast and adjust the exposure for low lighting images. Our first contribution is the modeling of image degradation in low lighting conditions. Then, the local maximum color value prior is proposed, i.e., in most regions of well exposed images, the local maximum color value of a pixel will be very high. By combining the image degradation model and local maximum color value prior, we propose to recover the un-degraded images under low lighting conditions. Last, an adaptive exposure adjustment module is proposed to obtain the final result. We show that our approach enables better enhancement comparing with popular image editing tools and academic algorithms.

Smart light random memory sprays Retinex: a fast Retinex implementation for high-quality brightness adjustment and color correction.

Removing the influence of illumination on image colors and adjusting the brightness across the scene are important image enhancement problems. This is achieved by applying adequate color constancy and brightness adjustment methods. One of the earliest models to deal with both of these problems was the Retinex theory. Some of the Retinex implementations tend to give high-quality results by performing local operations, but they are computationally relatively slow. One of the recent Retinex implementations is light random sprays Retinex (LRSR). In this paper, a new method is proposed for brightness adjustment and color correction that overcomes the main disadvantages of LRSR. There are three main contributions of this paper. First, a concept of memory sprays is proposed to reduce the number of LRSR's per-pixel operations to a constant regardless of the parameter values, thereby enabling a fast Retinex-based local image enhancement. Second, an effective remapping of image intensities is proposed that results in significantly higher quality. Third, the problem of LRSR's halo effect is significantly reduced by using an alternative illumination processing method. The proposed method enables a fast Retinex-based image enhancement by processing Retinex paths in a constant number of steps regardless of the path size. Due to the halo effect removal and remapping of the resulting intensities, the method outperforms many of the well-known image enhancement methods in terms of resulting image quality. The results are presented and discussed. It is shown that the proposed method outperforms most of the tested methods in terms of image brightness adjustment, color correction, and computational speed.

Enhancement of low exposure images via recursive histogram equalization algorithms

This paper proposes two exposure based recursive histogram equalization methods for image enhancement. The proposed methods are very effective for images acquired in low light condition like underwater sequences or night vision images. The first method is recursive exposure based sub-image histogram equalization (R-ESIHE) that recursively performs ESIHE [20] method till the exposure residue among successive iteration is less than a predefined threshold. The second method is named as recursively separated exposure based sub image histogram equalization (RS-ESIHE) that performs the separation of image histogram recursively; separate each new histogram further based on their respective exposure thresholds and equalize each sub histogram individually. The experimental results show that low exposure image enhancement problem was not addressed by earlier HE based methods, has been efficiently handled by these new methods. The performance evaluation of new methods is done in terms of image information content as well as visual quality inspection. The proposed methods outperforms earlier HE based contrast enhancement algorithms specifically for low light images.

Modified Sigmoid Function Based Gray Scale Image Contrast Enhancement Using Particle Swarm Optimization

The main objective of an image enhancement is to improve eminence by maximizing the information content in the test image. Conventional contrast enhancement techniques either often fails to produce reasonable results for a broad variety of low-contrast and high contrast images, or cannot be automatically applied to different images, because they are parameters dependent. Hence this paper introduces a novel hybrid image enhancement approach by taking both the local and global information of an image. In the present work, sigmoid function is being modified on the basis of contrast of the images. The gray image enhancement problem is treated as nonlinear optimization problem with several constraints and solved by particle swarm optimization. The entropy and edge information is included in the objective function as quality measure of an image. The effectiveness of modified sigmoid function based enhancement over conventional methods namely linear contrast stretching, histogram equalization, and adaptive histogram equalization are better revealed by the enhanced images and further validated by statistical analysis of these images.

Automatic selection of color constancy algorithms for dark image enhancement by fuzzy rule-based reasoning

This work introduces a fuzzy rule-based system operating as a selector of color constancy algorithms for the enhancement of dark images. In accordance with the actual content of an image, the system selects among three color constancy algorithms, the White-Patch, the Gray-World and the Gray-Edge. These algorithms have been considered because of their accurate remotion of the illuminant, besides showing an outstanding color enhancement on images. The design of the rule-based system is not a trivial task because several features are involved in the selection. Our proposal consists in a fuzzy system, modeling the decision process through simple rules. This approach can handle large amounts of information and is tolerant to ambiguity, while addressing the problem of dark image enhancement. The methodology consists in two main stages. Firstly, a training protocol determines the fuzzy rules, according to features computed from a subset of training images taken from the SFU Laboratory dataset. We choose carefully twelve image features for the formulation of the rules: seven color features, three texture descriptors, and two lighting-content descriptors. In the rules, the fuzzy sets are modeled using Gaussian membership functions. Secondly, experiments are carried out using Mamdani and Larsen fuzzy inferences. For a test image, a color constancy algorithm is selected according to the inference process and the rules previously defined. The results show that our method attains a high rate of correct selection of the most well-suited algorithm for the particular scene.

Frog and Haze Image Enhancement Using Improved Histogram Equalization Method in HIS Color Space

This paper concerns the problem of frog and haze image enhancement. Images are often degreed due to the fog and haze condition. In this paper, an image enhancement method by using improved histogram equalization in HIS color space was put forward. Firstly, the image was transformed from RGB to HIS color space. Then the S and I components were treated with improved histogram equalization separately. When judging whether a gray level was to be merged with another, the weight coefficients with increased step were assigned to these low frequency gray levels according to their distance to the current gray level. Thus the excessive gray level merging was avoided. At the same time, a non-linear gray level mapping algorithm was proposed, which improved the contrast and brightness of the image. The experimental results show that our methods could keep the original colors and details of the image better, and they could improve the frog and haze image display effects significantly.

Color images enhancement for edge information protection based on second order Taylor series expansion approximation

This article addresses the problem of color image enhancement for edge information protection. A new nonlinear adaptive enhancement (NNAE) algorithm is presented to resolve the problem in parallel procedure for low or high intensity and poor contrast (LIPC or HIPC) images. The NNAE algorithm consists of three steps. First, a RGB color image is converted to an intensity image, then an adaptive intensity adjustment with local contrast enhancement is performed parallelly, by using a single scale shift-variant Gaussian bilateral filter and the second order Taylor series expansion approximation technology, and finally colors are restored. A significant advantage of NNAE is that the edge information of enhanced images can be enhanced or preserved for LIPC or HIPC images. NNAE is only compared with Parallel Nonlinear Adaptive Enhancement (PNAE) algorithms. The experimental results show that the visual effects of NNAE are same or better than PNAE, and the enhanced images have more edges information changes by NNAE.

A parallel nonlinear adaptive enhancement algorithm for low- or high-intensity color images

Abstract This article addresses the problem of color image enhancement for images with low or high intensity and poor contrast (LIPC or HIPC). A parallel nonlinear adaptive enhancement (PNAE) algorithm using information from local neighborhood is presented to resolve the problem in parallel. The PNAE algorithm consists of three steps. First, a red-green-blue (RGB) color image is converted to an intensity image, then an adaptive intensity adjustment with local contrast enhancement is parallelly performed, and finally, colors are restored. The PNAE algorithm can be adjusted to control the level of enhancement on the overall lightness and the contrast achieved at the output separately. Most of the parameters used in PNAE are robust for LIPC and HIPC color image enhancement. Experimental results show that PNAE outperforms two popular methods in both computational efficiency and overall content preservation of image while improving local contrast for LIPC and HIPC image enhancement.

Implementation of cellular automata for impulsive noise reduction in grey scale images

This paper presents a new cellular automata based filter with the ability to remove impulsive noise, while, simultaneously, preserving edges and image details efficiently. Noise reduction is one of the most commonly used operations in image analysis. It is considered to be an important application of image processing; digital images can be corrupted by different types of noise during the image acquisition or transmission. Several noise reductions have been proposed in literature for enhancing the images. In this paper, a new and optimal approach of noise reduction based on cellular automata has been proposed. The idea is simple but effective technique for noise reduction that greatly improves the performances of complicated images. To demonstrate the capability of our filtering approach, it was tested on several different image enhancement problems. Results are compared with other existing filtering technique in terms of Peak Signal to Noise Ratio (PSNR). The comparative analysis of various image noise reduction methods is presented and shown that cellular automata based algorithm performs better than all these techniques under almost all scenarios.

Simple League Championship Algorithm

is a lot of discussion in the literature on the advantages of one soft computing technique over the other. Number of new meta heuristic algorithms are made available for solving different optimization problems. But always there is some ambiguity in deciding which technique is better than the other to solve a specific optimization problem in a particular context. A new methodology has been proposed in this paper which provides some theoretical frame work to use multiple optimization techniques at the same time onto a single optimization problem. This new approach has been implemented to solve image enhancement problem in a fingerprint. First the methodology of this new approach has been described then the application of the proposed methodology is presented with some experimental results.