scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Image Enhancement Algorithm Based on Depth Difference and Illumination Adjustment

17 Jul 2021-Scientific Programming (Hindawi)-Vol. 2021, pp 1-10
TL;DR: In this paper, a traffic image enhancement model based on illumination adjustment and depth of field difference is proposed to improve the clarity and color fidelity of traffic images under the complex environment of haze and uneven illumination and promote road traffic safety monitoring.
Abstract: In order to improve the clarity and color fidelity of traffic images under the complex environment of haze and uneven illumination and promote road traffic safety monitoring, a traffic image enhancement model based on illumination adjustment and depth of field difference is proposed. The algorithm is based on Retinex theory, uses dark channel principle to obtain image depth of the field, and uses spectral clustering algorithm to cluster image depth. After the subimages are divided, the local haze concentration is estimated according to the depth of field and the subimages are adaptively enhanced and fused. In addition, the illumination component is obtained by multiscale guided filtering to maintain the edge characteristics of the image, and the uneven illumination problem is solved by adjusting the curve function. The experimental results show that the proposed model can effectively enhance the uneven illumination and haze weather image in the traffic scene and the visual effect of the images is good. The generated image has rich details, improves the quality of traffic images, and can meet the needs of traffic practical application.

Content maybe subject to copyright    Report

Citations
More filters
Journal ArticleDOI
TL;DR: In this article , an ensembled spatial method for image enhancement was proposed, which employed the Laplacian filter, which highlights the areas of fast intensity variation, and then the gradient of the image was determined, which utilizes the surrounding pixels for the weighted convolution operation for noise diminishing.
Abstract: Most medical images are low in contrast because adequate details that may prove vital decisions are not visible to the naked eye. Also, due to the low-contrast nature of the image, it is not easily segmented because there is no significant change between the pixel values, which makes the gradient very small Hence, the contour cannot converge on the edges of the object. In this work, we have proposed an ensembled spatial method for image enhancement. In this ensembled approach, we first employed the Laplacian filter, which highlights the areas of fast intensity variation. This filter can determine the sufficient details of an image. The Laplacian filter will also improve those features having shrill disjointedness. Then, the gradient of the image has been determined, which utilizes the surrounding pixels for the weighted convolution operation for noise diminishing. However, in the gradient filter, there is one negative integer in the weighting. The intensity value of the middle pixel might be deducted from the surrounding pixels, to enlarge the difference between the head-to-head pixels for calculating the gradients. This is one of the reasons due to which the gradient filter is not entirely optimistic, which may be calculated in eight directions. Therefore, the averaging filter has been utilized, which is an effective filter for image enhancement. This approach does not rely on the values that are completely diverse from distinctive values in the surrounding due to which it recollects the details of the image. The proposed approach significantly showed the best performance on various images collected in dynamic environments.

4 citations

Journal ArticleDOI
TL;DR: The quality and diversity of the specific types of images generated by the proposed GAN are improved compared with the current mainstream GAN method with supervision, which is in line with the subjective evaluation results of human beings.
Abstract: Today, new media technology has widely penetrated art forms such as film and television, which has changed the way of visual expression in the new media environment. To better solve the problems of weak immersion, poor interaction, and low degree of simulation, the present work uses deep learning technology and virtual reality (VR) technology to optimize the film playing effect. Firstly, the optimized extremum median filter algorithm is used to optimize the “burr” phenomenon and a low compression ratio of the single video image. Secondly, the Generative Adversarial Network (GAN) in deep learning technology is used to enhance the data of the single video image. Finally, the decision tree algorithm and hierarchical clustering algorithm are used for the color enhancement of VR images. The experimental results show that the contrast of a single-frame image optimized by this system is 4.21, the entropy is 8.66, and the noise ratio is 145.1, which shows that this method can effectively adjust the contrast parameters to prevent the loss of details and reduce the dazzling intensity. The quality and diversity of the specific types of images generated by the proposed GAN are improved compared with the current mainstream GAN method with supervision, which is in line with the subjective evaluation results of human beings. The Frechet Inception Distance value is also significantly improved compared with Self-Attention Generative Adversarial Network. It shows that the sample generated by the proposed method has precise details and rich texture features. The proposed scheme provides a reference for optimizing the interactivity, immersion, and simulation of VR film.

3 citations

Proceedings ArticleDOI
14 Dec 2022
TL;DR: In this paper , a Super Resolution GAN (SRGAN) is used to super resolute the fine textures of the image by upscaling it and in order to enhance the images further, ESRGAN is used.
Abstract: There is tremendous amount of computational power in artificial intelligence models like computing variety of complex mathematical calculations and recognizing objects. In the past six to seven years, the amount of computing power used by record-breaking AI models doubled frequently in the time span of months. An interesting way in which these models learn and progress is through deep learning. Deep learning is an intelligent machine’s way in which machines learn without being supervised by us and grants them the power to recognize speech, translate, and even make or take data-driven decisions. Machines consider this as a studying method, inspired by the architecture of the human brain and how we learn. An important deep learning method where we train the machines on information that is unlabeled is called unsupervised learning. A strong part of neural networks that are utilized for unsupervised learning is Generative Adversarial Networks. When it comes to applications on images quality improvement, Super Resolution GAN (SRGAN) have a key role to play in it. It was proposed by researchers at Twitter. The motive of this GAN is to super resolute the fine textures of the image by upscaling it. In order to enhance the images further, ESRGAN is used. As the name suggests, ESRGAN is an implementation of SRGAN and uses some added components of SRGAN.

1 citations

Proceedings ArticleDOI
14 Dec 2022
TL;DR: In this article , a Super Resolution GAN (SRGAN) is used to super resolute the fine textures of the image by upscaling it and in order to enhance the images further, ESRGAN is used.
Abstract: There is tremendous amount of computational power in artificial intelligence models like computing variety of complex mathematical calculations and recognizing objects. In the past six to seven years, the amount of computing power used by record-breaking AI models doubled frequently in the time span of months. An interesting way in which these models learn and progress is through deep learning. Deep learning is an intelligent machine’s way in which machines learn without being supervised by us and grants them the power to recognize speech, translate, and even make or take data-driven decisions. Machines consider this as a studying method, inspired by the architecture of the human brain and how we learn. An important deep learning method where we train the machines on information that is unlabeled is called unsupervised learning. A strong part of neural networks that are utilized for unsupervised learning is Generative Adversarial Networks. When it comes to applications on images quality improvement, Super Resolution GAN (SRGAN) have a key role to play in it. It was proposed by researchers at Twitter. The motive of this GAN is to super resolute the fine textures of the image by upscaling it. In order to enhance the images further, ESRGAN is used. As the name suggests, ESRGAN is an implementation of SRGAN and uses some added components of SRGAN.
TL;DR: In this paper , the authors presented an approach for the segmentation and classification of brain tumors using Entropy and CLAHE (Contrast Limited Adaptive Histogram Equalization) based Intuitionistic Fuzzy Method with Deep Learning.
Abstract: The inner area of the human brain is where abnormal brain cells gather when they become a mass. These are known as brain tumors, and based on the location and size of the tumor, they can produce a wide range of symptoms. Accurate segmentation and classification of brain tumors are critical for effective diagnosis and treatment planning. In this paper, we present a novel approach for the segmentation and classification of brain tumors using Entropy and CLAHE Based Intuitionistic Fuzzy Method with Deep Learning. Entropy and CLAHE (Contrast Limited Adaptive Histogram Equalization) based Intuitionistic Fuzzy Method with Deep Learning is a technique that combines several image processing and machine learning algorithms to enhance the quality of images. By applying entropy-based techniques to an image, we can identify and highlight the most significant features or patterns in the image. Our study provides a thorough evaluation of the proposed technique and its performance compared to other methods, showing its effectiveness and potential for use in real-world applications. Our method separates the tumor regions from the healthy tissue and provides accurate results in comparison with traditional methods. The results of this study demonstrate the potential of this approach to improve the diagnosis and treatment of brain tumors and provide a foundation for future research in this field. The proposed technique holds significant promise for improving the prognosis and quality of life for patients with brain tumors.
References
More filters
Proceedings ArticleDOI
01 Aug 2016
TL;DR: Simulation shows that this proposed technique has better contrast enhancement, brightness preservation, hue preservation and better entropy, without gamut problem.
Abstract: The images, captured by camera, might suffer from poor contrast, saturation artefacts or improper brightness. Hence, image enhancement becomes an important step to improve the quality of image. Images are enhanced such that there is change in intensity or saturation component, keeping hue unchanged. Often, gamut problem arises when transforming from one plane to another. In this paper, the technique focusses on enhancing the contrast of low illumination images while at the same time maintaining the brightness. A function called exposure, is used in estimation of underexposed and highly underexposed images. The brightness of the input image is increased using transformation functions, which is then treated as target image so that the brightness of input image is adjusted close to target image, using histogram matching. The existing technique Exposure based Sub Image Histogram Equalization (ESIHE), is used to increase the visual quality, which is followed by guided image filter for edge smoothening. Simulation shows that this proposed technique has better contrast enhancement, brightness preservation, hue preservation and better entropy, without gamut problem.

4 citations

Book ChapterDOI
01 Jan 2020
TL;DR: Different methods have been used to get the accurate count of vehicles and their performances have been analyzed and popular image processing method background subtraction and deep learning algorithms: R-CNN, Fast R- CNN and Faster R-ESPN have been implemented.
Abstract: Traffic congestion has been an emerging issue when it comes to problems faced by commuters on road on a daily basis. It leads to loss of time, money, and fuel when one is stuck in a traffic jam. This has led to the need of more path-breaking technologies in the field of intelligent transport systems (ITS). Today, a lot of data are available which can be used to extract important information and perform the desired analysis. With CCTV surveillance cameras at almost every traffic pole, information like count of vehicles can be used to analyze the traffic patterns at a particular location. In this paper, different methods have been used to get the accurate count of vehicles and their performances have been analyzed. Popular image processing method background subtraction and deep learning algorithms: R-CNN, Fast R-CNN and Faster R-CNN have been implemented.

4 citations

Journal ArticleDOI
TL;DR: According to the experiments performed in the course of this work, the parallel MSR interpolation algorithm attained a speedup factor of 32 on two Intel MIC acceleration cards with 60 cores, which indicated that the parallel algorithm greatly reduced processing time and maximized speed and performance.
Abstract: Image‐enhancement algorithms, for example, median filtering algorithms, Gaussian filtering algorithms, and the multiscale Retinex (MSR) algorithm, are widely used in unmanned aerial vehicle (UAV) image processing to resolve the problems of poor clarity, insufficient contrast, and weak adaptability of the aerial images. Aiming to improve the low efficiency of processing a large volume of UAV images using the MSR algorithm, this research realized a parallel MSR algorithm using the OpenMP programming model based on Intel's many integrated core (MIC) platform. First, the principle and serial implementation of the MSR algorithm were reviewed in detail, and the algorithm's hotspots were determined with the Intel VTune tool. Then, the corresponding parallel algorithm was designed and implemented. After checking the correctness of the parallel algorithm, systematical experiments on UAV images of different sizes were carried out. According to the experiments performed in the course of this work, the parallel MSR interpolation algorithm attained a speedup factor of 32 on two Intel MIC acceleration cards with 60 cores, which indicated that the parallel algorithm greatly reduced processing time and maximized speed and performance.

4 citations

Journal ArticleDOI
TL;DR: In the original publication, Equations were incorrectly presented and the original article has been corrected.
Abstract: In the original publication, Equations were incorrectly presented. The original article has been corrected.

3 citations

Proceedings ArticleDOI
01 Dec 2018
TL;DR: The achieved results show the strength of the proposed art is better than already developed methods in FVR domain, including single scale retinex filter with chromaticity preserved algorithm and Gaussian filter to enhance the low and high quality finger vein images.
Abstract: As a secure and reliable biometric trait, finger vein recognition (FVR) can be employed to verify the individuals in real-time applications. However, the pattern of vein is unclear in some finger vein images due to light scattering by the skin and non-uniform illumination, which deteriorates the performance of the FVR system. To deal with the image quality problem, a novel finger-vein image quality assessment method and an enhancement method are proposed. The proposed FVR Scheme is based on two folds: (i) Image Quality Assessment, and (ii) Image Enhancement. First, the quality of the image is assessed by the decision tree with r-smote technique, to classify the finger vein image into two classes, i.e. High Quality (HQ) and Low Quality (LQ) images. Second, a single scale retinex filter (SSR) with chromaticity preserved algorithm and Gaussian filter are proposed to enhance the low and high quality finger vein images. Total of 1052 finger vein images are employed for the testing aspect of quality evaluation, enhancement and recognition method. After that, low error rate EER of 0.0379 is obtained by the proposed art. Finally, the achieved results show the strength of the proposed art is better than already developed methods in FVR domain.

3 citations