Introduction Image processing is a method to convert animage into digital form and perform some operations on it, in order to get anenhanced image or to extract some useful information from it. Image enhancement plays a fundamental role inimage processing applications where the experts make decision with respect tothe image information. Image enhancement means improvement of an imageappearance by increasing dominance of some features or by decreasing ambiguitybetween different regions of the image. 1.
Image enhancement techniquesImage enhancement techniques can beclassified into two methods:1. Spatial domain methods 2.Frequencydomain methods. Spatial domain method act on pixels directly.The pixel values are altered to achieve the desired enhancement. Pointprocessing methods , Gray Level Transformation, log transformation, histogramprocessing, Image Negatives ,morphological operators,Piecewise Linear Transformation,Global Power Law Transform, AdaptivePower Law Transform, Spatial Filtering arespatial domain enhancement methods.
Frequencydomain method is a term used to describe the analysis of mathematical functionsor signals with respect to frequency and operate directly on the imagetransform coefficients. Commonly used transform co-efficient are FourierTransform(FT), Discrete Cosine Transform (DCT), Discrete Wavelet Transform(DWT).The basic idea is to enhance the image by manipulating the transformcoefficients. In Frequency Domain method, the image is converted to frequencydomain. Hence the Fourier Transform of the image is computed first. All theenhancement operations are performed on the Fourier transform of the image andthen the Inverse Fourier transform is performed to get the desired result. 2.
Contrast Enhancement TechniquesContrastEnhancement aims at enhancing the global or local contrast of an image. Some ofthe enhancement techniques include histogram equalization, Genetic algorithms andfuzzy set algorithms. Many modifications are done in the standard methods toimprove the contrast of the images. 3.
Literature SurveyAn overview ofseveral contrast enhancement methods is discussed in the literature. Cheolkon Jung 1proposed a new contrast enhancement method named Optimized Perceptual ToneMapping (OPTM) which focuses on human visual attention. This method is a threestep process. Initially to measure the human visual attention, a saliencyhistogram is constructed.
Next the contrast enhancement of images subject toconstraints like maximum tone distortion is performed .At the end, to avoid overenhancement the pixel mapping function is adjusted. The proposed algorithmachieves improved performance and good results without over enhancement.
Butthe method requires more time for contrast enhancement compared to Histogram equalization(HE) and contrast limited histogram equalization (CLAHE) Anil Singh Parihar 2 proposed a fuzzydissimilarity histogram (FDHE) algorithm for contrast enhancement and extendedit to Fuzzy Contextual Contrast Enhancement (FCCE).In order to capture intensity level differences in the neighborhood of the pixelsFDHE is used.It provides global contrast enhancement. When the contrast wasincreased, the fuzzy theory helped in retaining the continuity in smoothimages. FDHE was extended to FCCE to provide a contextual intensity transform function. FCCE uses bothglobal and local enhancement. FDHE and FCCE involve no parameters.
Originalshape of the histogram is preserved. Shilpa Suresh 3 proposed a novel cuckoosearch based enhancement algorithm for the enhancement of satellite images. Theproposed method is implemented in three phases: a chaotic initialization phase,an adaptive levy flight strategy and mutative randomization phase. Thealgorithm improves the convergence rate of the standard CS algorithm and provesincreased adaptability for different images. This method shows littlecomplexity in execution. Huanjing Yue 4 proposed to enhance imagesby estimating illumination and reflectance layers through intrinsic imagedecomposition.
The split Bregman algorithm is adopted to solve thedecomposition problem. The Gamma correction is performed after decomposition toboost the details globally. Later CLAHE is used for the enhancement of localdetails.
Results show high performance than the other decomposition models. Theproposed method is designed only for contrast enhancement. It cannot be usedfor other methods like surface re-texturing, object insertion, and video enhancement. SeEunKim 5 proposed an entropy based contrast enhancement method in the waveletdomain. Initially it uses a local entropy scaling in the wavelet domain.
Forentropy scaling in the wavelet transform domain to enhance image contrast,mathematical works were used and then a color enhancing method in the HSI colorspace was developed. The algorithm worked in two steps: The low frequencycoefficients in the wavelet domain are modified and then the saturationcomponent of the HIS color space is linearly scaled by using the enhancedintensity component. By using the proposed method the details and colorinformation of low light images are good without any post processing. M.Shakeri 6proposed a contrast enhancement algorithm based on local histogram equalizationwhich was used for automatic determination of the number of sub-histograms anddensity based histogram division. The algorithm worked in three stages. Initially,the estimation of the number of clusters forimage brightness levels is done using histogram equalization.
In the nextstage, the image brightness levels areclustered and finally include the contrast enhancement for each individual cluster separately. Thealgorithm is compared with other methods based on quality and quantitymeasurement. Lalit Maurya 7 proposeda social spider optimization algorithm which produces two enhanced images onewith high contrast, increased entropy and the other image with increased peaksignal to noise ratio. Later the two enhanced images are combined to get aneffective image. Comparisons were done with HE, Linear contrast stretching,Standard Particle Swarm Optimization.
Results show that the proposed methodachieves high PSNR, brightness preservation and contrast enhancement of anygiven input image which leads to better visual quality JeyongShin 8 proposed histogram-based locality- preserving CE (HBLPCE), anoptimization problem to preserve the localities of the histogram for performingcontrast enhancement. By this method the shape of the enhanced image remainsthe same as the original image. The objective function of the optimizationproblem is formed to find a least squares solution of locality conditions. Theexperimental results show that HBLPCE adapts well on images with variousstatistical properties. HuangLidong 9 proposed an image enhancement method CLAHE-DWT which combines bothCLAHE and DWT. The algorithm works in three stages.
Initially the original imageis decomposed into low frequency and high frequency components by DWT. In thesecond stage, low frequency coefficients are enhanced using CLAHE and highfrequency coefficients are unchanged to limit the noise. Finally the image isreconstructed by taking the inverse DWT of the new coefficients. LEI, Noiseestimation, PSNR, MAE are the parameters used for evaluation.
Results showimpressive performance and the over enhancement could be avoided Anil Singh Parihar 10 proposes an entropy-baseddynamic sub-histogram equalization algorithm for contrast enhancement. Arecursive division of the histogram is performed based on the entropy of thesub histograms. It provides a better distribution of intensity levels over theentire dynamic range, which results in better contrast.
It preserves theoriginal characteristics of the image resulting in contrast-enhanced images. Parameterswere not used. Results were compared with conventional contrast enhancementalgorithms. DaeyeongKim 11 proposed an Adaptive Contrast Enhancement algorithm which is formulatedtopreserve the shape of the 1-D histogram and the statistical information on thegray-level differences betweenneighboring pixels obtained by a 2-D histogram.
The proposed methodworks in 2 stages. Initially, toenhance the entire contrast by stretching the 1-D histogram while preservingthe shape of the histogram. Then to improve the details of nonsmooth areasoccurring frequently in input images. Constrained optimization problemwas formulated and the enhanced images were obtained using quadraticprogramming. Experimental results show the enhanced images with good imagequality. Also the results show insufficient color quality and color root meanenhancement measure (CRME) KristoforB 12 proposed a new joint contrastenhancement and turbulence mitigation (CETM) method that utilizes estimationsfrom the contrast enhancement algorithm to improve the turbulence removalalgorithm. An analysis of fog and turbulence is incorporated in this method.
TurbulenceMitigation Metric (TMM) is also proposed to evaluate turbulence. It is observedthat removing fog before frame averaging is a better approach than removing fogafter frame averaging because of the depth discontinuities in scenes. Forremoving turbulence it is common to average the motion compensated images together inorder to remove the turbulent artifacts. MayankTiwari 13 proposed a highspeed quantile-based histogramequalisation (HSQHE) algorithm for contrast enhancement suitable for highcontrast digital images. HSQHE divides the input image histogram into two or more sub-histograms, wheresegmentation is based on quantile values and hence the entire spectrum of greylevel plays a vital role in enhancementprocess. The recursive segmentation of the histogram is not done, so only aminimal time is required for segmentation.For the Assessment of contrastenhancement PSNR, Entropy metrics are used.
For Assessment of brightnesspreservation AMBE is used. HSQHE preserves image brightness more accurately inless time interval ZhaoWei 14 proposed a EMHM(Entropymaximization histogram modification) method, which consists of dividing theglobal histogram equalization into two steps, pixel populations mergence (PPM)step and the grey-levels distribution (GLD) step .The histogram of input imageis merged with the proposed entropy maximisation rule (EMR) in the PPM step, which can minimizethe reduction of entropy because of mergence and compression of the number of the number of grey scalewith non-zero pixel populations(GNPP) in output histogram. In the GLD step, thenew grey levels are redistributed using a LDF, which can alleviate the contrastoverstretching.
Proposed method performs better than the existing methods. MohsenAbdoli 15 proposed a new contrast enhancement method named GMMCE (Gaussianmixture model-based contrast enhancement) to enhance low contrast images. This method models the histogram of low-contrastimage by the combination of a limited number of Gaussians where each Gaussianpresents a dominant intensity level of the image. This modelling attempts toreflect the shape of a narrow histogram in the parameters of individualGaussians, to convey it to a broadened version.
The global contrast enhancementof the image was achieved by the enhancement of sub-histograms separated by themean value of the Gaussians of the GMM. Experimental results show that theshape preserving method of GMMCE enhances the contrast of the image. S.
No Authors Title of the paper Methodology Features Advantages Disadvantages 1 Cheolkon Jung, Tingting Sun Optimized Perceptual Tone Mapping for Contrast Enhancement of Images Optimized Perceptual Tone Mapping (OPTM) Focuses on the human visual attention by constructing a saliency histogram and performs Contrast Enhancement Improves the performance without over enhancement Needs more time for CE compared to HE,CLAHE 2 Anil Singh Parihar, Om Prakash Verma, Chintan Khanna Fuzzy-Contextual Contrast Enhancement Fuzzy dissimilarity histogram (FDHE), Fuzzy Contextual Contrast Enhancement (FCCE) Captures the intensity level differences in the neighborhood of the pixels Global and local CE. No parameters are used. Original shape of histogram is preserved EME measure is low 3 Shilpa Suresh, Shyam Lal, Chintala Sudhakar Reddy, Mustafa Servet Kiran A Novel Adaptive Cuckoo Search Algorithm for Contrast Enhancement of Satellite Images Novel Adaptive cuckoo search based enhancement algorithm (ACSEA) Contrast enhancement for satellite images Improved convergence rate. Good efficiency and robustness Complex in its execution 4 Huanjing Yue, Jingyu Yang, Xiaoyan Sun, Feng Wu Contrast Enhancement Based on Intrinsic Image Decomposition Split Bregman algorithm and CLAHE To enhance images by estimating illumination and reflectance layers through intrinsic image decomposition Good enhancement Designed only for CE.
Cannot be used for methods like surface re-texturing, object insertion etc 5 Se EunKim,JongJu Jeon,IlKyuEom Image contrast enhancement using entropy scaling in wavelet domain An entropy based contrast enhancement method in the wavelet domain Used in HSI color space and performs image contrast enhancement Color information of low light images are good without any post processing. Over-enhanced regions exist 6 M.Shakeri, M.
H.Dezfoulian, H.Khotanlou, A.H.Barati, Y.Masoumi Image contrast enhancement using fuzzy clustering with adaptive cluster parameter and sub-histogram equalization Contrast enhancement algorithm based on local histogram equalization Determination of the number of sub-histograms and density based histogram division Natural appearance of images and enhanced the contrast Loss of details in high brightness levels of the image. Noise in the output image. 7 Lalit Maurya, Prasant Kumar Mahapatra, Amod Kumar A social spider optimized image fusion approach for contrast enhancement and brightness preservation A social spider optimization (SSO)algorithm Improvement in sharpness, PSNR, brightness preservation Better visual quality The number of edge pixels of HE technique is high while the fitness value is less 8 Jeyong Shin, Rae-Hong Park Histogram-Based Locality-Preserving Contrast Enhancement Histogram-based locality- preserving CE (HBLPCE) To preserve the localities of the histogram for performing contrast enhancement Adapts well on images with various statistical properties.
Longer Execution time of global CE for small images. 9 Huang Lidong, Zhao Wei , Wang Jun, Sun Zebin Combination of contrast limited adaptive histogram equalisation and discrete wavelet transform for image enhancement Combines both CLAHE and DWT To enhance the local details of an image Performs well in detail preservation and noise suppression. High-frequency component which contains most of the noise in original image is unchanged 10 Anil Singh Parihar , Om Prakash Verma Contrast enhancement using entropy-based dynamic sub-histogram equalization Entropy-based dynamic sub-histogram equalization algorithm (EDSHE) Performs a recursive division of the histogram based on the entropy of the sub histograms Natural-looking, good contrast images Entropy measure is less for few images. 11 Daeyeong Kim, Changick Kim Contrast Enhancement Using Combined 1-D and 2-D Histogram-Based Techniques Histogram stretching technique ,quadratic programming To preserve the shape of the 1-D histogram the statistical information on the gray-level differences Enhanced images and perceptual image quality Processing time is slower 12 Kristofor B.
Gibson and Truong Q. Nguyen An Analysis and Method for Contrast Enhancement Turbulence Mitigation Contrast enhancement and turbulence mitigation (CETM) Provides an analysis of fog turbulence Increased contrast Less time consumption PSNR is very low 13 Mayank Tiwari, Bhupendra Gupta, Manish Shrivastava High-speed quantile-based histogram equalisation for brightness preservation and contrast enhancement Highspeed quantile-based histogram equalisation (HSQHE) Contrast enhancement suitable for high contrast digital images Image brightness more accurately in less time interval High PSNR value only for certain images 14 Zhao Wei, Huang Lidong, Wang Jun, Sun Zebin Entropy maximisation histogram modification scheme for image enhancement Entropy maximization histogram modification (EMHM) Divides the global histogram equalization into two steps, pixel populations mergence (PPM) step and the grey-levels distribution (GLD) step Good enhancement, avoids amplified noise and image artefacts. Contrast Overstretching problem 15 Mohsen Abdoli, Hossein Sarikhani, Mohammad Ghanbari, Patrice Brault Gaussian mixture model-based contrast Enhancement Gaussian mixture model-based contrast enhancement)(GMMCE) Uses Gaussian mixture modeling of histograms to model the content of the images Lowest approximation error highest similarity to the original histogram low-complexity method Subjective Quality measures not used