Adaptive typhoon cloud image enhancement using genetic algorithm and non-linear gain operation in undecimated wavelet domain
Introduction
There are many kinds of noise in a typhoon cloud image. If they cannot be efficiently reduced, they may affect the overall image quality to the extent that it is impossible to extract some important information. For example, the noise may disturb the procedure of locating the center position, when we try to predict the moving path of the typhoon. In addition, the contrast of some typhoon cloud images may be poor, which may affect in accurately segmenting the helical cloud band from the typhoon cloud image. Therefore, it is important to efficiently reduce the noise and enhance the contrast in a typhoon cloud image. Recently, much good work has been carried out in satellite cloud image enhancement. Albertz and Zelianeos (1990) developed an approach by merging the data from several satellite images of the same area in order to improve the spatial resolution of the sensor. This enhancement technique is called Data Cumulation. The paper starts with the theory of sampling image data over a scene, discusses the theoretical background of the approach and describes its implementation. Simulated Data Cumulation has been carried out using both artificial targets and satellite image data as well. Fernandez-Maloigne pointed out that threshold selection is an important problem for the purpose of image enhancement. Image enhancement consists of subdividing the intensity gray levels into bands, such that the resulting image presents more contrast and less gray levels (Fernandez-Maloigne, 1990). One means of achieving threshold selection is through the use of gray tone spatial dependency matrices. A function of gray levels is computed from a co-occurence matrix of the image and then a threshold is chosen that corresponds to the etrema of this function. This article describes the algorithm of GTRLM and the results of the application of this technique to some satellite images. Bekkhoucha presented a method for the improvement of the visual quality of satellite and aerial images (Bekkhoucha and Smolarz, 1992). It helps geographers in their visual interpretation of urban growth in developing countries. This method is based on local contrast estimation. Results of the application of the algorithms to aerial and satellite images are discussed. Konstantinos accomplished satellite image enhancement and smoothing towards automatic feature extraction through an effective serial application of anisotropic diffusion processing and alternating sequential filtering (Konstantinos Karantzalos, 2004). A robust anisotropic diffusion filtering is used with Tukey’s biweight robust error norm for “edge-stopping” function. A well-known class of morphological filters, alternating sequential filtering, is applied afterwards for a more extended enhancement and smoothing.
Recently, the wavelet transform has been widely applied to image enhancement. Some image enhancement algorithms only consider the enhancement and do not care about noise reduction. For example, Fu and Wan used the improved histogram equalization in wavelet domain to enhance the image (Fu et al., 2000; Wan and Shi, 2007). Temizel proposed two algorithms to enhance an image resolution: by the estimation of detail wavelet coefficients at high-resolution scales and cycle-spinning methodology in the wavelet domain ([52], [53]). Shi and Derado, respectively, proposed two algorithms to enhance the contrast of an image by only modifying the detail coefficients in wavelet domain (Shi et al., 2006; Derado et al., 2007). Xiao presented an algorithm to enhance the contrast of an image by modifying both coarse and detail coefficients (Xiao and Ohya, 2007). Heric introduced a novel image enhancement technique based on the multiscale singularity detection with an adaptive threshold whose value is calculated via the maximum entropy measure in the directional wavelet domain (Heric and Potocnik, 2006). Scheunders proposed a wavelet-based enhancement method based on Bayesian estimation for multicomponent images or image series (Scheunders and De Backer, 2005). Some image enhancement algorithms only consider noise reduction and do not care about detail enhancement. For example, Ercelebi proposed a method by applying lifting-based wavelet domain Wiener filter to enhance the contrast of an image (Ercelebi and Koc, 2006). The proposed method transforms an image into the wavelet domain using lifting-based wavelet filters and then applies a Wiener filter in the wavelet domain and finally transforms the result into the spatial domain.
Many enhancement algorithms consider both detail enhancement and noise reduction. For example, Zeng proposed a wavelet-based algorithm for image contrast enhancement (Zeng et al., 2004). The approach treats the correlation between wavelet planes as an indication of the likelihood that noise is present. Then, it modifies the wavelet transform coefficients at different scales in different degrees by a pointwise non-linear transformation. The algorithm achieves an excellent balance between the enhancement of subtle image detail and the avoidance of noise amplification. Jung Claudio described a new method for noise suppression and edge enhancement in digital images based on the wavelet transform (Jung Claudio and Scharcanski, 2004). This method is adaptive to different amounts of noise in the image, and tends to be more robust to larger noise contamination than comparable techniques. Nakashizuka proposed a non-linear image enhancement filter for noisy images (Nakashizuka et al., 2004). In order to avoid the emphasis of the noises, weighted unsharp masking techniques have been proposed. In these methods, the high-frequency component is defined as a product between a weighting function of which modulus increases around image edges and the linear high-pass filter output. Sun, Zhou, and Yong proposed three image enhancement algorithms based on wavelet transform (Sun and Beom, 2005; Zhou et al., 2002; Yong et al., 2006). These algorithms can efficiently enhance the contrast of an image while suppressing speckle. Sakellaropoulos presented a method for mammographic image de-noising and contrast enhancement based on an overcomplete dyadic wavelet transform (Sakellaropoulos et al., 2002). De-noising is accomplished by adaptive soft-thresholding and contrast enhancement by a local non-linear gain operator. Sattar proposed a non-linear multiscale reconstruction method for image enhancement using dual-tree complex wavelet transform (Sattar and Gao, 2003). This image enhancement method reduces additive noise while preserving the sharpness of the image. Jung proposed a method for image de-noising with edge preservation and enhancement, based on image multi-resolution decomposition by a redundant wavelet transform (Jung and Scharcanski, 2003). Within the proposed framework, edge-related coefficients may be enhanced and de-noised simultaneously. Mencattini presented an algorithm for mammographic image enhancement and de-noising based on the wavelet transform and local iterative fuzzy noise variance estimation (Mencattini et al., 2005). Luo presented a new method of X-ray image de-noising based on fast lifting wavelet thresholding for general noise reduction and spatial filtering for further de-noising by using a derivative model to preserve edges (Luo et al., 2006). Belousov Artem described a developed two-phase full-color image enhancement algorithm (Belousov Artem et al., 2006). During the first phase the picture is de-noised-based wavelet thresholding. During the second phase brightness and contrast are automatically tuned up using evolutionary algorithm. Wu proposed an image enhancement method by combing partial differential equation (PDE) with wavelet shrinkage (Wu and Ruan, 2006). Compared with traditional smoothing models, the new hybrid model has no Gibbs phenomena, smoothes image without staircasing and enhances edge to preserve feature and texture of image.
Over the last decade, there has been much interest in discrete undecimated wavelet methods in signals and images. Therefore, hundreds of papers have been published in journals throughout the scientific and engineering disciplines. Recently, UWT has been widely applied to image processing, such as image de-noising (Gyaourova et al., 2002; Florian and Thierry, 2007; Alin et al., 2004; Gnanadurai and Sadasivam, 2005; Fabrizio et al., 2005; [10], [11]; Fabrizio and Luciano, 2002; Fabrizio and Gionatan, 2003; Chambolle and Lucier, 2001; Sveinsson and Atli Benediktsson, 2003; Sveinsson et al., 1998; Mahmoud and Ibraheem, 2003; Wang et al., 2003; Peng et al., 2004; Raghavendra and Bhat, 2006), image fusion (Liu et al., 2006; Styliani and Vassilia, 2007; Rockinger, 1997), image restoration (Fujiwara et al., 2006; Ciarlini and Cascio, 2005), image compression (Hui et al., 1997), image enhancement (Beaulieu et al., 2003; Wang et al., 2003; Lemeshewsky, 2002), image code (Liang and Thomas, 1998; Wu and Su, 1999; Al-Mohimeed and Mohammed, 1998), feature extraction (Pun and Lee, 2004) and segmentation (Zhang et al., 2006), etc. The quality of reconstructed image using UWT is better than classical discrete orthogonal wavelet transformation. It can get rid of the Gibbs phenomena when the de-noising image is obtained by inverse UWT.
Now some image enhancement methods based on wavelet transform consider only detail enhancement and not the noise reduction or suppression (Fu et al., 2000; Wan and Shi, 2007; Temizel and Vlachos, 2006; Temizel and Vlachos, 2005; Shi et al., 2006; Derado et al., 2007; Xiao and Ohya, 2007; Heric and Potocnik, 2006; Scheunders and De Backer, 2005). Some image enhancement methods consider only reducing noise and not detail enhancement (Ercelebi and Koc, 2006). Some image enhancement methods consider both noise reduction and detail enhancement. However, most of them estimate the de-noising threshold using the statistic properties of noise (Sun and Beom, 2005; Zhou et al., 2002; Yong et al., 2006; Sakellaropoulos et al., 2002; Jung and Scharcanski, 2003; Mencattini et al., 2005; Luo et al., 2006; Belousov Artem et al., 2006; Wu and Ruan, 2006). In fact, it is difficult because the accurate statistic properties of the noise cannot be priorly known or accurately predicted. Moreover, most of them enhance detail by user interference so as to obtain good results (Shi et al., 2006; Derado et al., 2007; Xiao and Ohya, 2007; Zeng et al., 2004; Sakellaropoulos et al., 2002). This will constrain their wide applications in actual image enhancement. In order to solve the above problems, an efficient enhancing algorithm for a typhoon cloud image is proposed by employing UWT and genetic algorithm (GA). First, UWT is implemented to a typhoon cloud image. Noise is then reduced in the fine high-frequency sub-bands of each decomposition level so that the maximum signal-noise ratio can be obtained in the high-frequency sub-bands. The de-noising threshold is estimated by genetic algorithm (GA) in the UWT domain. Detail is enhanced by an improved non-linear gain operator in the coarse high-frequency sub-bands of each decomposition level. In order to automatically obtain the non-linear gain parameters, GA is used to search for the optimal non-linear gain parameters. Experimental results show that the proposed algorithm can efficiently reduce the additive gauss white noise (GWN) while enhancing the detail in the typhoon cloud image. Finally, we compare our algorithm with other some similar image enhancing algorithms. Fig. 1 shows the diagram of the proposed enhancing algorithm, where UWT presents undecimated wavelet transform, IUWT indicates inverse undecimated wavelet transform.
Section snippets
Discrete undecimated wavelet transform
The undecimated discrete wavelet transform is redundant and shift invariant and it gives a dense approximation to the continuous wavelet transform than the approximation provided by the orthonormal discrete wavelet transform. From the filter bank point of view, we keep both even and odd downsamples, and further split the lowpass bands. We can also show the UWT from the matrix point of view. The undecimated discrete wavelet transform can be visualized as a matrix multiplicationwhere y is a 1×
De-noising principle with UWT
Wavelet transform of the colored noise is non-stationary. De-noising effect is not satisfactory by employing traditional “global threshold” to reduce noise in an image. Fortunately, Johnstone has proved that wavelet transform of the colored noise is still stationary at all scales of every resolution level (Johnstone and Silverman, 1997). Thus de-noising thresholds at all scales of every resolution level are calculated to effectively reduce noise in an image.
We consider discrete image model as
De-noising threshold estimation in UWT domain
According to Section 3, we have to minimize in order to obtain the asymptotic optimal de-noising threshold. Jansen used the “gold segmentation” method to obtain the de-noising threshold. However, this method is only efficient to the single peak function. In fact, curve is just like the shape in Fig. 2. Thus we cannot obtain good de-noising threshold only by using “gold segmentation”. We will use GA to resolve this problem.
GA can find the near-global optimal solutions in a large
Detail enahancement in UWT domain
Next, based on discrete UWT, a kind of non-linear enhancement operation, which was proposed by Laine in 1994, is employed to enhance the local contrast for image (Laine et al., 1994). We suppress pixel values of very small amplitude, and enhance only those pixels that are larger than a certain threshold Φ within each level of transform space. We design the following function to accomplish this non-linear operation:where
Enhanced image quality assessment
Digital images are subject to a wide variety of distortions during acquisition, processing, compression, storage, transmission and reproduction, any of which may result in a degradation of visual quality. For applications in which images are ultimately to be viewed by human beings, the only “correct” method of qualifying visual image quality is through subjective evaluation. In practice, however, subjective evaluation is usually too inconvenient, time-consuming and expensive (Wang et al., 2004
Experimental results
In experiments, three infrared typhoon cloud images :TAILIM (NO. 0513), SONAMU (NO. 0611) and CIMARON (NO. 0620), which are provided by China Meteorological Administration, China National Satellite Meteorological Center, are used to verify the efficiency of the proposed algorithm, the three typhoon cloud images are corrupted by additive gauss white noise (GWN). In order to demonstrate the efficiency of the proposed algorithm, we will compare the performance between the proposed algorithm (ZCJ),
Conclusion
In this paper, we propose an algorithm to suppress the noise and extrude detail for a typhoon cloud image by the GCV, non-linear gain operation with GA in UWT domain. GA is used to obtain the asymptotic optimal de-noising threshold in the UWT domain without the accurate statistic properties of noise priorly known. The optimal nonlinear gain parameters can adaptively be also obtained by GA in the UWT domain. An efficient objective assessment measure, which combines information entropy, contrast
Acknowledgments
Part of the research is supported by the Grants for Research Foundation of State Key Laboratory of Remote Sensing Science, Jointly Sponsored by the Institute of Remote Sensing Applications of Chinese Academy of Sciences and Beijing Normal University (2009KFJJ013). China Meteorological Administration, China National Satellite Meteorological Center is acknowledged for providing all the typhoon cloud images in this manuscript. Dr. C. J. Dumanmu is acknowledged for polishing the manuscript.
References (65)
- et al.
Enhancement of satellite image data by data cumulation
Journal of Photogrammetry and Remote Sensing
(1990) - et al.
Film grain reduction on colour images using undecimated wavelet transform
Image and Vision Computing
(2004) - et al.
Noise reduction in spine videofluoroscopic images using the undecimated wavelet transform
Computerized Medical Imaging and Graphics
(2004) - et al.
Wavelet-based histogram equalization enhancement of gastric sonogram images
Computerized Medical Imaging and Graphics
(2000) - et al.
Undecimated wavelet based speckle reduction for SAR images
Pattern Recognition Letters
(2005) Generalized cross validation for wavelet thresholding
Signal Processing
(1997)- et al.
A genetic algorithm approach to color image enhancement
Pattern Recognition
(1998) - et al.
Arbitrarily shaped image coding by using translation invariant wavelet transforms
Signal Processing
(1999) - et al.
E, Astrophysical image denoising using bivariate isotropic cauchy distributions in the undecimated wavelet domain
International Conference on Image Processing
(2004) - et al.
Li, Motion estimation and compensation based on almost shift-invariant wavelet transform for image sequence coding
International Journal of Imaging Systems and Technology
(1998)
Multi-Spectral Image Resolution Refinement Using Stationary Wavelet Transform
International Geoscience and Remote Sensing Symposium
Technique of images contrast enhancement: an application to satellite and aerial images
Automatique Productique Informatique Industrielle
Interpreting translation-invariant wavelet shrinkage as a new image smoothing scale space
IEEE Transactions on Image Processing
Wavelet image interpolation (WII): a wavelet-based approach to enhancement of digital mammography images
Lecture Notes in Computer Science
Lifting-based wavelet domain adaptive Wiener filter for image enhancement
IEE Proceedings Vision, Image and Signal Processing
Speckle suppression in ultrasonic images based on undecimated wavelets
Eurasip Journal on Applied Signal Processing
Speckle removal from SAR images in the undecimated wavelet domain
IEEE Transactions on Geoscience and Remote Sensing
Despeckling sar images in the undecimated wavelet domain: A map approach
IEEE International Conference on Acoustics, Speech, and Signal Processing
Defocused image restoration using translation invariant wavelet transform
SICE-ICASE International Joint Conference
Spatially adaptive wavelet thresholding with context modeling for image denoising
IEEE Trans on Image Processing
On the feasibility of cross-validation in image analysis
SIAM J.Appl. Math
Image enhancement by using directional wavelet transform
Journal of Computing and Information Technology—CIT
Image compression using shift-invariant dyadic wavelets
International Conference on Image Processing
Wavelet threshold estimators for data with correlated noise, Journal of the Royal Statistical Society
Series B
Adaptive image denoising and edge enhancement in scale-space using the wavelet transform
Pattern Recognition Letters
Wavelet transform approach to adaptive image denoising and enhancement
Journal of Electronic Imaging
Cited by (22)
Catenary image enhancement using wavelet-based contourlet transform with cycle translation
2014, OptikCitation Excerpt :In order to compare the practical results of different methods, we select a group of low contrast catenary images containing target, background and noise, which can test the performance such as the preservation of details and the suppression noise [9]. Meanwhile, in order to objectively assess the enhanced image quality [18]. We construct SNR and EPI to measure the enhancement effect.
Image contrast enhancement for preserving mean brightness without losing image features
2013, Engineering Applications of Artificial IntelligenceCitation Excerpt :In recent years, contrast enhancement techniques have been used in many digital multimedia systems, such as for object tracking in video surveillance systems (Chen et al., 2008), radiography applications (Dippel et al., 2002), arterial visualization of medical imaging (Bemmel et al., 2003), tumor microcirculation (Koh et al., 2003), virtual restoration of ancient Chinese paintings (Pei et al., 2004), vision impairment estimation (Tang et al., 2004), recovery of underwater visibility (Schechner and Karpel, 2005), bas-relief generation (Sun et al., 2009), and face recognition (Xie and Lam, 2005). Therefore, the aim of this study is to develop a high-performance contrast enhancement method that fits the current requirements (Perner et al., 1999; Shin et al., 1992; Zhang et al., 2010). Contrast enhancement methods can be broadly categorized into two major classes: direct and indirect methods (Arici et al., 2009).
Image enhancement based on wavelet transformation and pseudo-color coding with phase-modulated image density processing
2013, Infrared Physics and TechnologyCitation Excerpt :Adaptive typhoon cloud image enhancement using genetic algorithm and non-linear gain operation in undecimated wavelet domain has been proposed as an efficient enhancement algorithm for typhoon cloud image by combining discrete undecimated wavelet transform (UWT) with genetic algorithm (GA). The enhancement method can efficiently reduce the additive gauss white noise (GWN) while enhancing the detail in the typhoon cloud image [4]. A comparative study on thresholding methods in wavelet based image de-noising by Fei Xiao and Yungang Zhang explores the properties of several representative thresholding techniques in wavelets de-noising, considering both noise reduction and detail enhancement used by wavelet transform [10].
An intelligent noise reduction method for chaotic signals based on genetic algorithms and lifting wavelet transforms
2013, Information SciencesCitation Excerpt :While chaos-based approaches often are computationally expensive and may not be very effective either, especially when noise is strong [23]. In recent years wavelet transform (WT) has been widely applied to noise reduction [8,13,26,36,46]. It provides a trade-off between time and frequency resolutions and allows flexible exploitation of a signal at different resolution levels [27,28,30,37,41].
A Research on Typhoon Tracking System Based on Meteorological Remote Sensing
2020, 2020 International Conference on Internet of Things and Intelligent Applications, ITIA 2020