Finally, some examples of Retinex algorithm enhancement based on variational methods for low-light image enhancement are shown in Fig. Radio Eng China 36(11):5962, Nie R, He M, Cao J, Zhou D, Liang Z (2019) Pulse coupled neural network based mri image enhancement using classical visual receptive field for smarter mobile healthcare. Similar to WHE, Wong et al. In addition, the Mach zone phenomenon is explained through experiments and the rationality of the method is established. IEEE/CVF Conference on Computer Vision and Pattern Recognition 2018:32913300, Hu Y, He H, Xu C, Wang B, Lin S (2018) Exposure: a white-box photo post-processing framework. [63] proposed the MSR algorithm. However, there are also some disadvantages in the spatial domain image enhancement method, such as the lack of sufficient robustness and imperceptibility requirements [16]. There are some other evaluation indicators, such as average brightness (AB) [159], visual information fidelity (VIF) [160], tone mapped image quality index (TMQI) [161] and learned perceptual image patch similarity metric (LPIPS) [162]. Specifically, the decomposition part divides the image into a illumination image and reflected image, and the enhancement part is used to generate a high-quality image. Obviously, AHE is better than HE in enhancing the contrast of local details of the image, as shown by the red box in Fig. For high-frequency gray scales and low-frequency gray scales, the frequency is reduced, and for the gray scales in between, the frequency value is increased by gamma correction, and the total frequency is still 1. The detail enhancement effect of ABMHE algorithm is better than that of POSHE algorithm, but it needs to increase the prior calculation process of image area. Pattern Recognit Lett 36:1014, Land EH (1977) The retinex theory of color vision. IEEE Trans Image Process 27(4):20492062, Lv F, Li Y, Lu F (2019) Attention guided low-light image enhancement with a large scale low-light simulation dataset, arXiv: 1908.00682, Zamir SW, Arora A, Khan S,Hayat M, Khan FS, Yang M-H, Shao L (2020) Learning enriched features for real image restoration and enhancement, arXiv preprint arXiv:2003.06792, Zhu M, Pan P, Chen W, Yang Y (2020) Eemefn: Low-light image enhancement via edge-enhanced multi-exposure fusion network. [24] proposed adjacent-blocks-based modification for local histogram equalization (ABMHE), which divides the image into active area, inactive area and general area in advance according to the ratio of the value of the image gradient value, so that local areas in the image can be selected in a targeted manner. The . We first survey the unsupervised image enhancement methods, including histogram specification, Retinex model, deep learning and visual cortex neural network. Google Scholar, Polesel A, Ramponi G, Mathews V (2000) Image enhancement via adaptive unsharp masking. 2 in this paper. Cooper et al. 6e. Image enhancement refers to the process of highlighting certain information of an image, as well as weakening or removing any unnecessary information according to specific needs. [86] proposed a low-light image enhancement (LIME) method via illumination map estimation. Lv et al. Wang and Luo [65] improved Jobsons [9] method and proposed a multi-layer image enhancement model. df are the histograms of (ac), respectively. [149] proposed a deep learning method for low-light images enhancement. 10, \({L_{12}}\) and \({L_{23}}\) are weak connections among different sub-cells of HRYNN. In the work of Qi et al., the linking item \({L_{12}}\) of RYNN1 and RYNN2 is set as follows: where \({\beta _{12}}\) is the local linking item between RYNN1 and RYNN2: \(\beta _{12} = max (I)/mean(I)\). IEEE Trans Neural Netw 10(3):584590, Lindblad T, Kinser JM (1999) Inherent features of wavelets and pulse coupled networks. Siam J Imaging Sci 8(2):787826, Frankle JA, Mccann JJ (1983) Method and apparatus for lightness imaging, Provenzi E, Fierro M, Rizzi A, De Carli L, Gadia D, Marini D (2007) Random spray retinex: a new retinex implementation to investigate the local properties of the model. Some examples of image enhancement. The main aim of image enhancementis to enhance thequality andvisual appearance of animage, or toprovide a better transformrepresentationfor future automated image processing. IEEE Trans on Image Process 29:72177232, Eckhorn R, Reitboeck H, Arndt M, Dicke P (2014) Feature linking via synchronization among distributed assemblies: simulations of results from cat visual cortex. 691700, Aly HA, Dubois E (2005) Image up-sampling using total-variation regularization with a new observation model. IEEE Trans Neural Netw 20(12):19801986, Zhan K, Shi J, Wang H, Xie Y, Li Q (2017) Computational mechanisms of pulse-coupled neural networks: a comprehensive review. \\ \;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;(m = 1,2,3) \end{array} \end{aligned} \end{aligned}$$, $$\begin{aligned}&Z = {Z_1} + {Z_2} + {Z_3} \end{aligned}$$, $$\begin{aligned} {L_{12}} = {\beta _{12}}\sum {{W_{ijkl}}(1,2){Z_{1,kl}}(n - 1)} \end{aligned}$$, $$\begin{aligned} {L_{23}} = {\beta _{23}}\sum {{W_{ijkl}}(2,3){Z_{2,kl}}(n - 1)} \end{aligned}$$, $$\begin{aligned} {W_{ijkl}}(1,2) = {W_{ijkl}}(2,3) = \frac{1}{{{{(i - k)}^2} + {{(j - l)}^2}}} \end{aligned}$$, \(\mathop x\limits ^ \sim = F(\mathop y\limits ^ \sim ) = F \circ G(x)\), $$\begin{aligned} {{\mathcal {L}}}_{content}= \frac{1}{{{C_j}{H_j}{W_j}}}\left\| {{\varPsi _j}(x) - {\varPsi _j}(\mathop x\limits ^ \sim )} \right\| \end{aligned}$$, $$\begin{aligned} {{\mathcal {L}}}_{color}= & {} - \sum \limits _i {\log {D_c}(G{{(x)}_b})} \end{aligned}$$, $$\begin{aligned} {{\mathcal {L}}}_{texture}= & {} - \sum \limits _i {\log {D_t}(G{{(x)}_g})} \end{aligned}$$, $$\begin{aligned} {{\mathcal {L}}}_{tv} = \frac{1}{{CHW}}\left\| {{\nabla _x}G(x) + {\nabla _y}G(x)} \right\| \end{aligned}$$, $$\begin{aligned} AMBE(X,Y) = \left| {MB(X) - MB(Y)} \right| \end{aligned}$$, $$\begin{aligned} SSIM(x, y) = l{(x,y)^\alpha } \cdot c{(x,y)^\beta } \cdot s{(x,y)^\gamma } \end{aligned}$$, $$\begin{aligned} l(x,y)&= \frac{{2{\mu _x}{\mu _y} + {c_1}}}{{\mu _x^2 + \mu _y^2 + {c_1}}},\quad c(x,y) = \frac{{2{\sigma _x}{\sigma _y} + {c_2}}}{{\sigma _x^2 + \sigma _y^2 + {c_2}}},\nonumber \\ s(x,y)&= \frac{{{\sigma _{xy}} + {c_3}}}{{\sigma _x^{}\sigma _y^{} + {c_3}}} \end{aligned}$$, \(\mathrm{{SSIM(x,y)}} \in \mathrm{{(0,1)}}\), $$\begin{aligned} PSNR = 10{\log _{10}}(\frac{{peakval{^2}}}{{MSE}}) \end{aligned}$$, $$\begin{aligned} MSE = \frac{1}{{mn}}\sum \limits _{i = 0}^{m - 1} {\sum \limits _{j = 0}^{n - 1} {{{\left\| {I(i,j) - {I_0}(i,j)} \right\| }^2}} } \end{aligned}$$, $$\begin{aligned} DE(X) = - \sum \limits _{i = 0}^{255} {p({x_i})\log p({x_i})} \end{aligned}$$, $$\begin{aligned} EME(X) = \frac{1}{{{k_1}{k_2}}}\sum \limits _{i = 1}^{{k_1}} {\sum \limits _{j = 1}^{{k_2}} {20In\frac{{\max ({X_{ij}})}}{{\min ({X_{ij}})}}} } \end{aligned}$$, $$\begin{aligned} L(x,y) = \mathop {\max }\limits _{c \in \{ r,g,b\} } {I^c}(x,y) \end{aligned}$$, $$\begin{aligned} RD(x,y) = \nonumber \\ \sum \limits _{i = 1}^m {\sum \limits _{j = 1}^n {(U(L(x,y),L(i,j)) \oplus U({L_e}(x,y),{L_e}(i,j)))} } \end{aligned}$$, $$\begin{aligned} LOE = \frac{1}{{m*n}}\sum \limits _{i = 1}^m {\sum \limits _{j = 1}^m {RD(i,j)} .} Each image enhancement method has its own advantages and disadvantages. Specially, a neural network is trained to generated a set of pixel-wise operators, which can simultaneously predict the noise and the illumination layer in the bilateral space. The Retinex-like model has the disadvantage of a large calculation. Pattern Recognit Lett 104:1522, Zheng WT, Pu T, Cheng J, Zheng H (2012) Image contrast enhancement by contourlet transform and pcnn, In: International conference on audio, language and image processing 735739, Singh GMA (2014) Various image enhancement techniques-a critical review. IEEE Trans Image Process 19(11):28252837, Ma W,Morel JM, Osher S, Chien A (2011) An l1-based variational model for retinex theory and its application to medical images, In: The 24th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2011, Colorado Springs, CO, USA, 20-25 June 2011, Osher S, Burger M, Goldfarb D, Xu J, Yin W (2005) An iterative regularization method for total variation-based image restoration. Various image enhancement algorithms were mentioned and underlying difficulties, limitations, merits and disadvantages were discussed in applying these techniques in the past two decades with three aspects: supervised algorithm, unsupervised algorithm and quality evaluation, respectively. In addition, due to the lack of ground truth, the paper proposes an adaptive perceptual loss including global and local to constrain the feature difference between the low-light input image and its enhanced image, and together with the adversarial loss used to train EnlightenGAN to form the final loss. In addition, the perception of color has a strong correlation with reflectivity [47]. The purpose of our review is to provide relevant researchers with a comprehensive and systematic analysis on image enhancement techniques and give them a valuable reference. The input receptive field \({F^S}\) can enhance the sharp edge information of images [118] and connection receptive field \({F^I}\) with different forms of orientation connection has a key role in the orientation tuning response of image processing. The spiking cortical model (SCM) is a variant model of PCNN, which was proposed by Zhan et al. In addition, W(1,2) and W(2,3) are weak linking synaptic weight and set as follows: The parameter \({\alpha _e}\) is the exponential decay coefficient of dynamic threshold E. \(F_m^I\)(m = 1,2,3) represent the orientation connection of \({30^ \circ },{90^ \circ }\) and \({120^ \circ }\), respectively, which detect input stimuli in different directions. Image enhancement is one of the most important technologies in the field of image processing, and its purpose is to improve the quality of images for specific applications. Infrared Phys Technol 86:5965, Lamberti F, Montrucchio B, Sanna A (2006) Cmbfhe: a novel contrast enhancement technique based on cascaded multistep binomial filtering histogram equalization. In 1974, Horn et al. I-SECRET: Importance-Guided Fundus Image Enhancement via Semi Magnetic resonance imaging (MRI) technique provides important information for doctors to diagnose diseases. In addition, \({c_1}\), \({c_2}\) and \({c_3}\) are small constants, just to avoid the fact that the denominator is zero. Some examples of Retinex algorithm enhancement based on center/surround are shown in Fig. In addition, \(\alpha\), \(\beta\) and \(\gamma\) are parameters employed to adjust the relative importance of the three components. Firstly, most of the predictions are performed under a low-resolution bilateral grid; Secondly, the entire neural network structure learns an affine transformation; Thirdly, most of the operations are performed at low resolution, but the loss function is finally established on the original resolution, in other words, the original resolution image can be optimized based on low resolution operations. Based on the excellent feature extraction capabilities of deep convolutional neural networks, many methods for low-brightness image enhancement based on multi-level features fusion have emerged. J Ambient Intell Hum Comput 10(10):40594070, Johnson JL, Padgett ML (1999) Pcnn models and applications. Google Scholar, Arici T, Dikbas S, Altunbasak Y (2009) A histogram modification framework and its application for image contrast enhancement. Image processing is the process of transforming an image into a digital form and performing certain operations to get some useful information from it. In recent years, researchers have employed the gamma function to modify the existing cumulative distribution function and implement the histogram frequency weighting technique [33, 34]. I is the input image. The image features easier by removing noise and other artifacts in a picture. IEEE Trans Image Process 13(4):600612, Shannon CE (2001) A mathematical theory of communication. IEEE Trans Consum Electron 53(2):757764, Wong CY, Liu S, Liu SC, Rahman A, Lin SCF, Jiang G, Kwok NM, Shi H (2016) Image contrast enhancement using histogram equalization with maximum intensity coverage. PubMedGoogle Scholar. The global item D(I) is used to control the final generated image and satisfies the GW (Gray world) assumption [76]. In 2018, Li et al. Therefore, Rahman et al. Magn Reson Med 54(6):14391447, Chen ZY, Abidi BR, Page DL, Abidi MA (2006) Gray-level grouping (glg): an automatic method for optimized image contrast enhancementpart i: the basic method., IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society 15(8) 2290302, Sheikh HR, Bovik AC (2006) Image information and visual quality. In addition to the above enhancement algorithm, Zhan et al. In addition, each sub-problem is transformed into a vectorized least squares regression, with closed-form solutions. In addition to the unsupervised image enhancement methods mentioned above, In this section, we will mainly introduce several weakly-supervised enhancement methods for low-light enhancement. In Sect. A example of visual comparison based on different image enhancement is shown in Fig. Neural Comput 2(3):293307, Reitboeck HJ (1989) A model for feature linking via correlated neural activity, Synergetics on Cognition, Parodi O, Combe P, Ducom JC (1996) Temporal coding in vision: coding by the spike arrival times leads to oscillations in the case of moving targets. IEEE Trans Consum Electron 49(4):13101319, Wang Zhou, Bovik AC, Sheikh HR, Simoncelli EP (2004) Image quality assessment: from error visibility to structural similarity. 543554, Sun Z, Diao M, Zhao Z (2006) Algorithm of texture image enhancement based on pcnn. Based on Rybaks theory, Qi et al. Note that EME is highly sensitive to noise. or less than j, and T is the total number of pixels.The main purpose of histogram equalization is to find gray level transformation function T to transform image f such that the histogram of T(f) is equalized. 12 that AHE and SRIE have better performance in detail enhancement, but HE, WTHE and AGCWD cause serious loss of details, such as the edge of a teacup, etc. Early literature [99] shows that the brightness perceived by the human visual system is different from the brightness of the light source, and it has a logarithmic relationship with the illuminance obtained by the human eye. Wang et al. where f, g denote the attenuation time constant and d is a positive constant for the globally inhibitory. 140 questions with answers in IMAGE ENHANCEMENT | Science topic Image enhancement is the process of digitally manipulating a stored image using software. The input \({S_{ij}}\) is only included in the feedback input \({F_{ij}}\). As mentioned above, based on the mechanism of human visual characteristics, the time matrix of PCNN can be used as an image enhancement tool. Although the special issue focuses on medical image enhancement, other important image processing methods are reported from submitted papers, including image reconstruction, enhancement, segmentation, feature extraction, and recognition. [25] proposed a cascaded multistep binomial filtering histogram equalization (CMBFHE) method based on the POSHE algorithm. Optik Int J Light Electron Opti 27(1):407414, Liu L, Jia Z, Yang J, Kasabov N (2015) A medical image enhancement method using adaptive thresholding in nsct domain combined unsharp masking. 63066314, Jiang Y, Gong X, Liu D, Cheng Y, Fang C, Shen X, Yang J, Zhou P, Wang Z (2019)Enlightengan: Deep light enhancement without paired supervision, arXiv preprint arXiv:1906.06972, Xiong W, Liu D, Shen X, Fang C, Luo J (2020) Unsupervised real-world low-light image enhancement with decoupled networks, arXiv preprint arXiv:2005.02818, Yang W, Wang S, Fang Y, Wang Y, Liu J (2020) From fidelity to perceptual quality: A semi-supervised approach for low-light image enhancement, In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. IEEE Trans Pattern Anal Mach Intell 28(9):15191524, Ng MK, Wang W (2011) A total variation model for retinex. In (46), f is a nonlinear threshold function. But in general, it is desirable to conduct both quantitative and qualitative assessments. Obviously, HRYNN has the best detail preservation performance, as marked by red boxes. Ma and Osher [81] introduced a total variation (TV) noise suppression model [82] to estimate the image illumination component. IEEE Trans Image Process 27(2):938948, Provenzi E (2017) Formalizations of the retinex model and its variants with variational principles and partial differential equations. Appl Sci Technol 10:58, Dengchao F, Zhaoxuan Y, Zengmin W (2007) Adaptive enhancement algorithm of color image based on improved pcnn, In: 2007 8th International Conference on Electronic Measurement and Instruments, IEEE, 2007, pp. Detailed algorithm content, which can be found in the original papers, are missed. \(\sigma\) represents kernel parameter, which can determine the enhancement effect of image. MathSciNet [27] proposed a partially overlapped sub-block histogram equalization (POSHE), which is the most classic local histogram equalization algorithm. The image processing technique begins with image pre-processing to improve the quality of the original image. What is Image Enhancement? - Definition from Techopedia Neural Netw 61:85117, Shen D, Wu G, Suk H-I (2017) Deep learning in medical image analysis. IEEE Trans on Neural Netw 22(6):880892, Ranganath H, Kuntimad G, Johnson J (1995) Pulse coupled neural networks for image processing. The evaluated image enhancement methods cover BIMEF [164], LIME [86], MSRCR [10], MF [163], SRIE [85], Dong [165], NPE [157] , DHECI [166], BPDHE [40], HE, Ying [168], WAHE [31], JED [167], Robust [87], LLNet [171], DeepUPE [169], GLADNet [170], MBLLEN [139] and Lv [141]. In Rybaks theory, \({F^S}\) is defined as a local connection. Image processing shows an important part of collecting information on brain images. Importance of Image Enhancement Techniques in Color Image Segmentation: A Comprehensive and Comparative Study Dibya Jyoti Bora Color image segmentation is a very emerging research topic in the area of color image analysis and pattern recognition. [127] proposed a unpaired learning method based on GAN. 7. J Electron Imaging 25(6):061603, LeCun Y, Bengio Y, Hinton G (2015) Deep learning. [54] used a double helix method to select the path. IEEE Computer Society Pr, Chang D, Wu W (1998) Image contrast enhancement based on a histogram transformation of local standard deviation. Image enhancement is important because of its usefulness in virtually all image processing applications. In 2009, Zhan et al. U(x,y) is step functions, and \(\oplus\) represent the exclusive OR operation. [127]s method has the advantages of stable performance, fast speed, and outstanding effects for image enhancement, it also has many potential shortcomings. In addition, supervised methods such as LLNet [171], DeepUPE [169], GLADNet [170], MBLLEN [139] and Lv [141] basically achieved a better performance. a is the original image. The Alternating Direction Multiplier Method (ADMM) [84] is used to simultaneously estimate the illumination and reflectance components. [145] thought that the traditional multi-scale Retinex (MSR) [63] algorithm can be regarded as a feedforward convolutional neural network with different Gaussian convolution kernels. Comput Eng Appl 39(19):9395, Kinser JM, Lindblad T (1999) Implementation of pulse-coupled neural networks in a cnaps environment. For the basic PCNN model, the parameters are set as follows: That is, the synapse weight between neurons is defined as the inverse of the square of the Euclidean distance [96, 101, 102]. The foremost purpose of the image enhancement is to get a clearer picture of the image which will further used in image processing technologies. IEEE Trans Consum Electron 53(4):17521758, Zhao W, Xu Z, Zhao J, Zhao F, Han X (2014) Infrared image detail enhancement based on the gradient field specification. [83] proposed a probabilistic model for image enhancement (PIE). Typical operations include filtering with morphological operators, histogram-based equalization, brightness, and contrast adjustment (Acharya and Ray, 2005 ). For (3), they separated the illumination component B(x,y) and the reflection component R(x,y) from L(x,y) and constructed the partial differential equations. The loss function is the F-norm square of the error matrix with regular terms. In recent years, related works in large numbers have taken noise as an important factor of Retinex-like, and proposed some low-light image enhancement models. For example, eliminating noise, revealing blurred details, and adjusting levels to highlight features of an image. Importance of Image Enhancement in Graphic Design Neural Netw 4(1):313, Chen Y, Park S-K, Ma Y, Ala R (2011) A new automatic parameter setting method of a simplified pcnn for image segmentation. This study, based on the proposed method of average gradient expression, improves the contrast of objects and enhances the edge information pertaining to objects relative to other methods and . Generally, we first normalize the histogram and then perform subsequent processing. Therefore, some researchers began to work on weakly supervised or unsupervised deep learning methods for low-light image enhancement. Multiscale Model Simul 4(2):460489, Palmaamestoy R, Provenzi E, Bertalmio M, Caselles V (2009) A perceptually inspired variational framework for color enhancement. \({c_1}\) and \({c_2}\) are the spatial distribution of the excitement intensity and inhibitory intensity of the receptive field respectively. Image enhancement plays an important role in improving image quality in the field of image processing, which is achieved by highlighting useful information and suppressing redundant information in the image. According to Land [62], it is considered that the Retinex result of a certain point of the image should be the ratio of the luminous intensity of that point to the average luminous intensity of the adjacent positions. Fundus images are widely used for eye disease screening [5, 8].However, the fundus image quality is inhomogeneous, and low-quality ones may severely affect clinical judgments and decisions [3, 16, 18, 19].Therefore, it is of great significance to enhance fundus images. To solve this problem, this paper presents a new and effective method for image enlargement and enhancement based on adaptive inverse hyperbolic tangent . Although the Retinex algorithms based on center/surround have a good effect in terms of color fidelity, the time complexity is high, the halo phenomenon is obvious and image distortion is inevitable. [90] proposed a new prior constraint, ie., local flatness. [150] proposed a deep convolutional neural network (CNN) to predict the HDR values of images. In 2018, about 187,000 new articles that investigated image enhancement issues can be found on Google Scholar. It is usually necessary to consider combining a variety of simple and effective algorithms, or fusing other enhancement techniques to achieve the final enhancement effect, so comprehensive image enhancement technology will also become a the key of follow-up study. In Gaberiel et al.s work, a large dataset of HDR images is collected to train CNN. Fu et al. IEEE Trans Consum Electron 49(4):13101319, Tan TL, Sim KS, Tso CP (2012) Image enhancement using background brightness preserving histogram equalisation. [151] proposed a Deep Reciprocating HDR Transformation (DRHT). 11. J Electron Imaging 27(1):011003, Horn BK (1974) Determining lightness from an image. For example, eliminating noise, revealing blurred details, and adjusting levels to highlight features of an image. In their method, two-way generative adversarial networks (GANs) with improvement are employed as the baseline. Image enhancement: contrast enhancement, part I Image enhancement techniques have been widely used in many applications of image processing where the subjective quality of images is important for human interpretation. IEEE Trans Circuits Syst Video Technol 11(4):475484, Agarwal M, Mahajan R (2017) Medical images contrast enhancement using quad weighted histogram equalization with adaptive gama correction and homomorphic filtering. \end{aligned}$$, $$\begin{aligned} {U_{ij}}\left[ n \right]= & {} f{U_{ij}}\left[ {n - 1} \right] + \left( {\sum \limits _{kl} {{M_{kl}}{Y_{kl}}\left[ {n - 1} \right] } + {S_{ij}}} \right) \nonumber \\\times & {} \left( {1 + \beta \left( {\sum \limits _{kl} {{M_{kl}}{Y_{kl}}\left[ {n - 1} \right] } - d} \right) } \right) \nonumber \\ {E_{ij}}\left[ n \right]= & {} g{E_{ij}}\left[ {n - 1} \right] + {V_E}{Y_{ij}}\left[ {n - 1} \right] \nonumber \\ {Y_{ij}}[n]= & {} \left\{ \begin{array}{l} 1,{\quad } if{\quad } {U_{ij}}[n] > {E_{ij}}[n] \\ 0,{\quad } else \end{array} \right. Image enhancement related researches have a significant growth in the past few years. \end{aligned}$$, $$\begin{aligned} &{F^{S}}(x,y) = {a_1}\sqrt{{x^2} + {y^2}} {\rm exp} ( - \frac{{{x^2} + {y^2}}}{{{c_1}}}) \\ &- {a_2}\sqrt{{x^2} + {y^2}} \exp ( - \frac{{{x^2} + {y^2}}}{{{c_2}}}) \end{aligned}$$, $$\begin{aligned}&{X_{Sij}}[n] = {V_S}\sum \limits _{kl} {F_{ijkl}^S} {S_{ij}} \end{aligned}$$, $$\begin{aligned} X_m^I[n]= & {} \alpha \sum \limits _{kl} {F_{m,kl}^I{Z_m}[n - 1]} \nonumber \\&\quad + {S_{ij}}\quad (m = 1,2,3) \end{aligned}$$, $$\begin{aligned}&{E_{m,ij}}[n] = {e^{ - {\alpha _e}}}{E_{m,ij}}[n - 1] \nonumber \\&\quad + {V_E}{Z_{m,ij}}[n - 1]\quad (m = 1,2,3) \end{aligned}$$, $$\begin{aligned}&{P_1}[n] = X_{ij}^S[n] - \frac{1}{{\phi }}X_1^I[n] + h \end{aligned}$$, $$\begin{aligned}&{P_2}[n] = X_{ij}^S[n] - \frac{1}{{\phi }}X_2^I[n] + h + {L_{12}} \end{aligned}$$, $$\begin{aligned}&{P_3}[n] = X_{ij}^S[n] - \frac{1}{{\phi }}X_3^I[n] + h + {L_{23}} \end{aligned}$$, $$\begin{aligned}&\begin{aligned} \begin{array}{l} {Z_{m,ij}}[n] = \left\{ \begin{array}{l} {k_m},\;\;\;\;if\;{P_{m,ij}}[n] > {E_{m,ij}}[n - 1]\\ 0,\;\;\;\;\;else \end{array} \right. Retinex model assumes that image lightness depends on illumination and re ectance. in Fig. Types of Image Processing In addition, Shi et al. [32] proposed a contrast enhancement method using a weighted histogram equalization (WHE), which weights the gray value of the HE result and the current gray value setting ratio to get the final result, so that the HE result only occupies a certain proportion. Image enhancement based on histogram modification.a is the original image. In recent years, a large number of Retinex-like models have been also emerging. In their work, the input image is first denoised by PCNN, and then segmented by PCNN, and the continuous output of multiple binary images is logically OR and exclusive or (EOR) operations to obtain image texture, edge and other information. In Qi et al.s work, input receptive field \({F_{ijkl}^S}\) is refined as follows: where \({a_1}\) and \({a_2}\) are the distribution sensitivity. The block diagram of the whole framework is shown in Fig. IEEE Trans on Image Process 27(6):28282841, Ren X, Yang W, Cheng W, Liu J (2020) Lr3m: robust low-light enhancement via low-rank regularized retinex model. As we know, Rybak neural networks are analogous to the components of actual neurons. For a grayscale image, a small Gaussian kernel can increase the brightness of the dark parts of the image, but it may cause excessive compensation and make the dynamic range of the image too small, ie., halo phenomenon. Anyone you share the following link with will be able to read this content: Sorry, a shareable link is not currently available for this article. Proc Vldb Endow 5(7):622633, DAndrade RG (1978) U-statistic hierarchical clustering. First of all, according to literature [153], the absolute mean brightness error(AMBE) is defined as: where MB(X) and MB(Y) represent the mean brightness of original image X and enhanced image Y, respectively. Image enhancement tools are often classified into (a) point operations, and (b) spatial operators. The visual cortex neural networks were proposed to simulate the mammalian visual cortex system. Importance of Image Enhancement and CDF for Fault Assessment of Photovoltaic Module Using IR Thermal Image. Then the relative brightness value at x can be obtained along this path: where n is the number of pixels on the path and \({\delta _\varepsilon }\) is threshold function. Although their method has the advantages of easy training and short time-consuming, it shows great limitations in enhancing low-quality images with noise, because the training set only uses high-quality images. [125] proposed a Weakly Supervised Photo Enhancer (WESPE) for low-light image enhancement based on Generative Adversarial Network (GAN). Image enhancement tools are often classified into (a) point operations, and (b) spatial operators. IEEE Trans Neural Netw 10(3):5918, Ranganath HS, Kuntimad G (1996) Iterative segmentation using pulse-coupled neural networks, In: Applications and Science of Artificial Neural Networks II, Vol. What is the importance of image enhancement? - Studybuff.com In image enhancement, in order to effectively eliminate the halo, scholars proposed the Retinex algorithm based on the variational framework, also known as the Retinex-like model. [139] proposed a multi-branch low-light enhancement network (MBLLEN). However, lots of researches [120, 121] have shown that the visual cortex system of mammals is heterogeneous.
امکان ارسال دیدگاه وجود ندارد!