Ratio rule and homomorphic filter for enhancement of digital colour image

doi:10.1016/j.neucom.2005.07.003

Neurocomputing

Volume 69, Issues 7–9, March 2006, Pages 954-958

https://doi.org/10.1016/j.neucom.2005.07.003 Get rights and content

Abstract

Digital colour image enhancement using a homomorphic filter provides good dynamic range compression, but it fails in colour rendition. In this letter, we propose to perform natural colour rendition in a digital colour image that is enhanced by a homomorphic filter. A novel neural network learning algorithm, named Ratio rule, is used to carry out the natural colour rendition process. The Ratio rule learns to restore colour in an image by representing the colour relationships of each pixel in the original image as a line of attraction in the state space. The dynamics of the neural network is then used as an associative memory for recalling the natural colour characteristics of image pixels after homomorphic filtering. Several experiments on benchmark problems show that the new technique of image enhancement can generate natural colour in images from their original colour information.

Introduction

High dynamic range is common in scenes that include both dark shadows and bright light sources. These scenes are difficult to represent because their dynamic range greatly exceeds the range of most electronic capturing devices. As a result, the dynamic range of an electronic image is compressed, obscuring actual scenic details and/or colours. Image enhancement is a process to improve the appearance of an electronic image as perceived by humans [7], [8] or to render the image more suitable for machine analysis [12]. The majority of enhancement algorithms have focused on the improvement of grey-level images [2], [4]. But, direct extensions of these processes into the colour domain have resulted in degraded or suboptimal images. Although it is possible to enhance a digital colour image by applying existing grey-level image enhancement algorithms to each red, green, and blue (RGB) component [2], [3], [7], [8], [13], the resulting image may not be enhanced optimally. Image distortion will result if the RGB components are not properly recombined after the grey-level enhancement algorithms are applied to each component.

In this letter, we propose a new neural network based technique to provide simultaneous improvement of the dynamic range and colour rendition of digital colour images. The method is based on a homomorphic processing system [3], [13] and a novel Ratio rule learning algorithm [10], [11]. The Ratio rule forms a line in the state space using all the stable points representing patterns of a family. All instances of this family with variations due to extraneous influence would converge to the line of attraction. The colour image enhancement process with a description of the line attractor is presented in Section 2. Section 3 provides the test results and discussions of the enhancement process and Section 4 presents the conclusions.

Section snippets

Colour image enhancement model

The human visual system can detect the range of light spectrum with wavelengths from 380 to 780 nm. Our perception of a colour is determined by which combination of R, G, and B sensors are excited and by how much they are excited [1]. Fig. 1 shows the spectral sensitivity of a typical human visual system. All digital colour devices that handle the storage and reproduction of colour images do the same by storing RGB values. Digitally storing an image requires that it first be broken down into a

Results and discussions

Several experiments have been conducted to validate the performance of the proposed method. The criterion for evaluation of the enhancement algorithm is judged by the dynamic range compression and the colour rendition of the image. The proposed method has been used to perform digital colour image enhancement in more than 100 images. These images are selected specifically for testing the dynamic range compression and colour restoration properties of the proposed method. Fig. 3 shows a typical

Conclusions

A new technique of colour image enhancement using a recurrent neural network in conjunction with a homomorphic filter is presented in this paper. The proposed method provides simultaneous improvement of the dynamic range and colour rendition to digital colour images. Moreover, it gives a new framework for applying any grey-level enhancement algorithm in the colour domain. This paper also demonstrates the use of an associative memory that encapsulates instances of colours for digital colour

Ming-Jung Seow was born in Kuala Lumpur, Malaysia, in 1979. He received his B.S. and M.S. degrees in Computer Engineering from the Old Dominion University, Virginia, USA, in 2001 and 2002, respectively. He is currently pursuing his doctoral degree in the Department of Electrical and Computer Engineering at the Old Dominion University. His research interests are in the areas of recurrent neural network architectures and learning algorithms, and applications of neural networks and machine

References (13)

P. Buser et al.
Vision
(1992)
V. Caselles et al.
Shape preserving local histogram modification
IEEE Trans. Image Process.
(1999)
O.D. Faugeras
Digital color image processing within the framework of a human visual model
IEEE Trans. Acoust. Speech Signal Process.
(1979)
R.C. Gonzalez
Digital Image Processing
(2002)
J.J. Hopfield
Neural networks and physical systems with emergent collective computational abilities
Proc. Natl Acad. Sci.
(1982)
S. Jankowski et al.
Complex-valued multi-state neural network associative memory
IEEE Trans. Neural Network
(1996)

There are more references available in the full text version of this article.

Cited by (165)

RepDehazeNet: Dual subnets image dehazing network based on structural re-parameterization
2024, Computers and Graphics (Pergamon)
In recent times, there has been notable and swift advancement in the field of image dehazing. Several deep learning techniques have demonstrated remarkable proficiency in resolving homogeneous dehazing issues. Nonetheless, the current dehazing approaches are generally formulated to deal with homogeneous haze, which is often undermined in real-world scenarios due to the uncertain haze dispersion. In this paper, we propose a dehazing model named RepDehazeNet by combining a structurally Reparameterization Encoder-Decoder subnet and a Full-Resolution Attention subnet. To be specific, the structural reparameterization idea is introduced into the encoder–decoder subnet to strengthen the feature extraction of dehazed images and improve the feature extraction speed. RepDehazeNet is compared with seven SOTA models on different datasets in terms of PSNR, SSIM, parameter quantity, and inference time. Compared to the DW-GAN model, the proposed RepDehazeNet model reduces the number of parameters by 2.7 million, and improves the inference speed by 90.3%, while achieving a higher PSNR of 0.5 dB on the NH-Haze2021 dataset. The experimental results demonstrate that the proposed RepDehazeNet model can effectively improve the real-time performance, accuracy of dehazing synthesized and nonhomogeneous haze images.
A new fast and efficient dehazing and defogging algorithm for single remote sensing images
2024, Signal Processing
Information about the earth’s surface is difficult to capture in remote sensing images because bad weather greatly curtails visibility and diminishes visual contrast in the images. For the purpose of military survey and aerial surveillance, these images are crucial for providing information. It is quintessential to eradicate bad weather conditions like haze and fog from remote sensing images. Recently, numerous deweathering initiatives and endeavors have been undertaken to alleviate these limitations. These contemporary deweathering approaches, however, are inadequate to recover dense haze images. This work presented a novel visibility restoration approach based on segmentation and unsharp mask guided filtering method. It consists of the following steps to restore the scene’s radiance: First, a segmentation method is employed to determine atmospheric light to quantify contrast and color. Estimation of the transmission map utilizing the dark channel prior is then performed to precisely determine the proximity between objects. The resultant output image contains halo artifacts and inconsistencies in the structure. A guided filter method based on unsharp masking is employed to optimize the transmission map in order to solve this issue. The experimental results demonstrate that the proposed indicators ensures high uniformity regarding qualitative and quantitative evaluation using six performance metrics: mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), universal image quality index (UQI), fog-aware density evaluator (FADE), dehazing algorithm index (DHQI), standard deviation $(C C)$ , and blind contrast enhancement assessment ( $e$ and $r$ ).
Single image fast dehazing based on haze density classification prior
2023, Expert Systems with Applications
Most of existing the dehazing methods have high computational complexity and poor dehazing quality. Therefore, a fast haze removal method based on the haze density classification prior is proposed. Haze density is reflected by the difference between maximum channel and minimum channel. Based on this, haze density prior (HDP) is proposed. The HDP can effectively distinction between mist image and dense haze image, and quickly estimate the atmospheric light veil. In addition, an optimized method for estimating atmospheric light using the mid-channel of haze image is proposed, which overcomes limitations of global atmospheric light. Our method can fast and efficiently recover clear image because there is no estimation of transmission map. Experiments show that the proposed method outperforms existing methods in haze removal, especially for dense haze images. Comparison of objective evaluation and running time shows that the proposed method is effective and real-time.
Haze removal for single image: A comprehensive review
2023, Neurocomputing
Image dehazing is always a hot topic in the field of computer vision since haze has significant impact on the imaging quality of camera. Therefore, many image dehazing methods have been proposed for the past decades. To help researchers who are new to this field quickly figure out the development history as well as the current status of image dehazing, this review analyzes some representative dehazing methods, evaluates their advantages and disadvantages, and most importantly, points out the best dehazing method from different viewpoints. A large quantity of experiments show that AECR-Net can be generally considered to be the best dehazing algorithm and Tarel method can be regarded as the best real-time dehazing method. Besides, the mainstream benchmark, metrics, challenges and opportunities for image dehazing are also discussed in this review.
Image dehazing via enhancement, restoration, and fusion: A survey
2022, Information Fusion
Citation Excerpt :
From such a perspective, the haze-degraded images can be enhanced by boosting the high-frequency component and reducing the low-frequency component. The commonly adopted approaches mainly include homomorphic filtering methods [46,47], wavelet transform methods [48,49] and curvelet transform methods [50]. Among those methods, the wavelet transform is more popular due to its efficiency and better ability to express the image in multi-resolution [169].
Haze usually causes severe interference to image visibility. Such degradation on images troubles both human observers and computer vision systems. To seek high-quality images from degraded ones, a large number of image dehazing algorithms have been proposed from different perspectives like image enhancement, restoration, and fusion. Especially in recent years, with the rapid development of deep learning, CNN-based methods have already dominated the mainstream of image dehazing and gained significant progress on benchmark datasets. This paper firstly presents a comprehensive survey of existing image dehazing methods, and then conducts both qualitative and quantitative comparisons among representative methods, from classic methods to recent advanced approaches. We expect the literature survey and benchmark analysis could help readers better understand the advantages and limitations of existing dehazing methods. Moreover, a discussion on possible trends in single image dehazing is put forward to innovate further works.
Improved dark channel priori single image defogging technique using image segmentation and joint filtering
2024, Science Progress

View all citing articles on Scopus

Vijayan K. Asari joined the Department of Electrical and Computer Engineering, Old Dominion University, Virginia as an Associate Professor in August 2000. Dr. Asari received the B.Sc. (Engg.) degree in Electronics and Communication Engineering from the University of Kerala, M.Tech. and Ph.D. degrees in Electrical Engineering from the Indian Institute of Technology, Madras, India. He has been working as an Assistant Professor at the T.K.M. College of Engineering (University of Kerala), India. He was with the National University of Singapore from 1996 to 1998 and with the School of Computer Engineering at Nanyang Technological University, Singapore from 1998 to 2000. Dr. Asari is the Director of the Computational Intelligence and Machine Vision Laboratory at ODU. His research activities are in the areas of neural networks, image processing, computer vision, and digital system architectures for application specific integrated circuits. Dr. Asari is a senior member of the IEEE.

View full text

LettersRatio rule and homomorphic filter for enhancement of digital colour image

Abstract

Introduction

Section snippets

Colour image enhancement model

Results and discussions

Conclusions

Vision

Shape preserving local histogram modification

IEEE Trans. Image Process.

Digital color image processing within the framework of a human visual model

IEEE Trans. Acoust. Speech Signal Process.

Digital Image Processing

Neural networks and physical systems with emergent collective computational abilities

Proc. Natl Acad. Sci.

Complex-valued multi-state neural network associative memory

IEEE Trans. Neural Network

Letters
Ratio rule and homomorphic filter for enhancement of digital colour image