9
368 JOURNAL OF DISPLAY TECHNOLOGY, VOL. 12, NO. 4, APRIL 2016 A Power-Saving Histogram Adjustment Algorithm for OLED-Oriented Contrast Enhancement Li-Ming Jan, Fan-Chieh Cheng, Chia-Hua Chang, Shanq-Jang Ruan, Senior Member, IEEE, and Chung-An Shen, Member, IEEE Abstract—For the modern multimedia devices, display resolu- tion and image quality are actively improved nowadays. Although such improvement can produce the high visual perception for the observer, the power consumption becomes an inevasible problem as it is rising progressively. In order to achieve a good balance be- tween visual perception and power consumption, we propose a his- togram-based power saving algorithm to improve the image con- trast for OLED display panels. The proposed algorithm modifies the empty bins of the image histogram as a pre-process of power reduction. Furthermore, the visual effect is compensated using the power saving histogram equalization algorithm. Experimental re- sults show that the proposed algorithm not only decreases the dis- play power to be lower than that of compared algorithms, but also generates the highly perceptual contrast of the images. Index Terms—Emissive display, contrast enhancement, power saving, histogram equalization. I. INTRODUCTION S MART multimedia devices have been quickly developed over the past decade and could be found in many appli- cations, such as medicine [1], vehicle detection [2], [4], and face recognition [3]. Fig. 1 shows the global output chart of flat panel displays (FPD) [5] used in multimedia devices during 2009 to 2013, based on the data from the Industrial Technology Research Institute of Taiwan. The thin film transistor liquid crystal display (TFT-LCD) is a representative non-emissive FPD technique, which produces very bright images and con- sumes less power than traditional cathode ray tube (CRT) monitors. However, power consumption is still a noticeable issue of the TFT-LCD. For example, TFT-LCD consumes 20–50% power rate in all embedded portable systems [6]. On the other hand, emissive displays are extensively used for wearable smart devices, and the organic light emitting diode (OLED) becomes the most popular emissive display technique in the next generation tendency, due to its properties of rich bright, full color, wide viewing angle, high contrast, and lower power consumption. In general, OLED can be classified as the passive matrix OLED (PMOLED) and the active matrix OLED (AMOLED), while the PMOLED is only suited for small-size Manuscript received August 02, 2015; revised September 14, 2015; accepted October 09, 2015. Date of publication October 26, 2015; date of current version March 10, 2016. This work was supported by the National Science Council under Grant NSC 101-2221-E-011-164-MY3. The authors are with the Department of Electronic and Computer Engi- neering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JDT.2015.2491998 Fig. 1. Global flat panel display output ratio. panel and required high pulse currents that leads pixel life span decreased. Thus, the AMOLED that can extend pixel life and lead to high efficiency is mainly utilized in the field. In smart multimedia devices, visual quality and power saving are two critical issues for display panels. In order to improve the quality of the captured images, contrast enhancement is a widely used pre-process and the histogram equalization (HE) is a well-known algorithm to be employed [7]–[18]. Celik et al. [7] built the normalized 2-D histogram based on the objective func- tion to enhance the contrast of the image, and [8] prposed to use Gaussian Mixture Model (GMM) to improve the visual quality of images. Furthermore, Ghita et al. [9] used the total variation (TV) minimization at a fidelity term to enhenace the image texture component, and Lee et al. [17] applied the optimiza- tion function through the 2-D histogram and reconstructed the output image by the mapping function with weighting vector. Moreover, Xu et al. [18] amalgamated contrast enhancement and white balancing into the combined structure of the input image histogram to achieve a balanced trade-off. In addition, fo- cusing on reducing the power consumption of TFT-LCD panels, the work in [10] utilized multi-histogram with the pixel distri- bution to overcome the image distortion for dynamic backlight dimming, whereas in [11], the suitable backlight current is ap- plied using the image histogram. Cho et al. in [12] considered image pixel compensation and clipping artifacts and Li et al. [13] provided the partitioned light guide (PLG) for backlight module structures. However, we note here that these works are targeting on the TFT-LCD panels, while the methods presented in this paper is focusing on OLED displays. Due to the funda- mental differences between these two technologies, the power reduction approaches proposed in [10]–[13] cannot be applied to OELD panels and vice versa. 1551-319X © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

368 JOURNALOFDISPLAYTECHNOLOGY,VOL.12,NO.4,APRIL2016 ...download.xuebalib.com/xuebalib.com.29556.pdf · 368 JOURNALOFDISPLAYTECHNOLOGY,VOL.12,NO.4,APRIL2016 APower-SavingHistogramAdjustmentAlgorithm

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

  • 368 JOURNAL OF DISPLAY TECHNOLOGY, VOL. 12, NO. 4, APRIL 2016

    A Power-Saving Histogram Adjustment Algorithmfor OLED-Oriented Contrast Enhancement

    Li-Ming Jan, Fan-Chieh Cheng, Chia-Hua Chang, Shanq-Jang Ruan, Senior Member, IEEE, andChung-An Shen, Member, IEEE

    Abstract—For the modern multimedia devices, display resolu-tion and image quality are actively improved nowadays. Althoughsuch improvement can produce the high visual perception for theobserver, the power consumption becomes an inevasible problemas it is rising progressively. In order to achieve a good balance be-tween visual perception and power consumption, we propose a his-togram-based power saving algorithm to improve the image con-trast for OLED display panels. The proposed algorithm modifiesthe empty bins of the image histogram as a pre-process of powerreduction. Furthermore, the visual effect is compensated using thepower saving histogram equalization algorithm. Experimental re-sults show that the proposed algorithm not only decreases the dis-play power to be lower than that of compared algorithms, but alsogenerates the highly perceptual contrast of the images.

    Index Terms—Emissive display, contrast enhancement, powersaving, histogram equalization.

    I. INTRODUCTION

    S MART multimedia devices have been quickly developedover the past decade and could be found in many appli-cations, such as medicine [1], vehicle detection [2], [4], andface recognition [3]. Fig. 1 shows the global output chart of flatpanel displays (FPD) [5] used in multimedia devices during2009 to 2013, based on the data from the Industrial TechnologyResearch Institute of Taiwan. The thin film transistor liquidcrystal display (TFT-LCD) is a representative non-emissiveFPD technique, which produces very bright images and con-sumes less power than traditional cathode ray tube (CRT)monitors. However, power consumption is still a noticeableissue of the TFT-LCD. For example, TFT-LCD consumes20–50% power rate in all embedded portable systems [6].On the other hand, emissive displays are extensively used forwearable smart devices, and the organic light emitting diode(OLED) becomes the most popular emissive display techniquein the next generation tendency, due to its properties of richbright, full color, wide viewing angle, high contrast, and lowerpower consumption. In general, OLED can be classified as thepassive matrix OLED (PMOLED) and the active matrix OLED(AMOLED), while the PMOLED is only suited for small-size

    Manuscript received August 02, 2015; revised September 14, 2015; acceptedOctober 09, 2015. Date of publication October 26, 2015; date of current versionMarch 10, 2016. This work was supported by the National Science Councilunder Grant NSC 101-2221-E-011-164-MY3.The authors are with the Department of Electronic and Computer Engi-

    neering, National Taiwan University of Science and Technology, Taipei 10607,Taiwan (e-mail: [email protected]).Color versions of one or more of the figures in this paper are available online

    at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/JDT.2015.2491998

    Fig. 1. Global flat panel display output ratio.

    panel and required high pulse currents that leads pixel life spandecreased. Thus, the AMOLED that can extend pixel life andlead to high efficiency is mainly utilized in the field.In smart multimedia devices, visual quality and power saving

    are two critical issues for display panels. In order to improvethe quality of the captured images, contrast enhancement is awidely used pre-process and the histogram equalization (HE) isa well-known algorithm to be employed [7]–[18]. Celik et al. [7]built the normalized 2-D histogram based on the objective func-tion to enhance the contrast of the image, and [8] prposed to useGaussian Mixture Model (GMM) to improve the visual qualityof images. Furthermore, Ghita et al. [9] used the total variation(TV) minimization at a fidelity term to enhenace the imagetexture component, and Lee et al. [17] applied the optimiza-tion function through the 2-D histogram and reconstructed theoutput image by the mapping function with weighting vector.Moreover, Xu et al. [18] amalgamated contrast enhancementand white balancing into the combined structure of the inputimage histogram to achieve a balanced trade-off. In addition, fo-cusing on reducing the power consumption of TFT-LCD panels,the work in [10] utilized multi-histogram with the pixel distri-bution to overcome the image distortion for dynamic backlightdimming, whereas in [11], the suitable backlight current is ap-plied using the image histogram. Cho et al. in [12] consideredimage pixel compensation and clipping artifacts and Li et al.[13] provided the partitioned light guide (PLG) for backlightmodule structures. However, we note here that these works aretargeting on the TFT-LCD panels, while the methods presentedin this paper is focusing on OLED displays. Due to the funda-mental differences between these two technologies, the powerreduction approaches proposed in [10]–[13] cannot be appliedto OELD panels and vice versa.

    1551-319X © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

  • JAN et al.: POWER-SAVING HISTOGRAM ADJUSTMENT ALGORITHM FOR OLED-ORIENTED CONTRAST ENHANCEMENT 369

    It is challenging to achieve a balance between power-savingand image contrast on emissive displays, since it does nothave the backlight module and tne power consumption isrelated to image pixels. Although the approach utilized in [14]mitigated this problem, the convex optimization function em-ployed therein needs to decide the minimal power consumptionconsisting with better contrast enhancement. For reducing thepower consumption of color images displayed on the emissivepanels, a histogram shrinking algorithm is proposed beforeenhancing the image contrast [16]. This method reduces theempty bins of the image histogram and keeps the image entropyconstant. Hence, it would be applicable to decrease the powerconsumption of emissive display and preserve the image detailssimultaneously. Although most of the previous techniques im-proved the image contrast and focused on the time complexityissues, the display power consumption has not been effectivelyreduced. While the TFT-LCD has a lot of advantages thantraditional CRT monitors, the promising smart device tech-nology makes emissive displays more potential ability to beemployed extensively. Furthermore, the TFT-LCD panel needsthe backlight module to support the display luminance which ispower-hungry and causes the panel volume lager than emissivedisplays. This paper proposes a novel method to deal withthis significant issue. We use the pre-process power-savinghistogram adjustment (PSHA) approach for darking imageluminance and keep the image details with entropy preserva-tion. Besides, the PSHA algorithm has a low time complexitycharacteristic, which suits for real-time operation.The organization of this paper is as follows: Section II de-

    scribes the power model used for contrast enhancement andthe compared emissive display-oriented HE-based methods.In Section III, the proposed method is described in detail.Section IV shows experimental results. Finally, we concludethis paper in Section V.

    II. RELATED WORKS

    Based on the emissive display power characterization ofpixels, the power of pixel level with OLED module has beenpresented in [19]. Let be the static power of OLED and bethe parameter of gamma correction. The power consumption ofa single-color pixel in an OLED panel can be formulated as

    (1)

    where and stand for the RGB values, andrepresent the weighting coefficients relied on specific displaypannel characteristic. In this model, gamma correction makesthe image color more convinced for human perception, whichis typically set as 2.2.However, the static power of OLED and RGB weighting co-

    efficients are independent with specific characteristic. Hence,we ignore the static power consumption and weighting coeffi-cient with RGB channels. Moreover, the total dissipated power(TDP) could be expressed with the gray level image by

    (2)

    where stands for each gray level pixel value and is thetotal number of image pixels. Obviously, the luminance of thecolor image dominates the total dissipated power in the abovemodel. Hence, the pixel value needs to be decreased whenperforming contrast enhancement. Based on this concept, thePower-Constrained Contrast Enhancement (PCCE) [14] andHistogram Shrinking (HS) [16] algorithms are disclosed andwill be briefly described in the following subsections.

    A. PCCE Method [14]The PCCE method is derived from the HE and logarithm

    function where the main concept of the HE is to generate theuniform image histogram. The histogram probability distribu-tion could re-allocate. The probability density function (PDF)is written as

    (3)

    where represents the number of pixels to the gray levelis the input image size, and is the highest intensity.

    The cumulative distribution function (CDF) is then expressedas

    (4)

    and the transformation function can be formulated by

    (5)

    where represents the desired transformation function. Inorder to employ the specific solution of power constrained andcontrast enhancement methods with pixel values for emissivedisplays, the calculation needs to include power and contrastterms. Hence, the PCCE method can re-formulates by (5). Thetransformation function is expressed as vector form and can bere-written with the differential matrix as

    ......

    .... . .

    ......

    (6)

    (7)

    where the is the normalized histogram. The histogram couldbe formulated as

    (8)

    where is the input image histogram. The PCCE algorithmemployed the histogram modification technique based on log-arithm function (LHM) for contrast enhancement, which couldbe neither controllable nor automatic.The LHM technique can be presented by the modi-

    fied histogram with the vector form , where. The modified transformation

    function is then expressed by

    (9)

  • 370 JOURNAL OF DISPLAY TECHNOLOGY, VOL. 12, NO. 4, APRIL 2016

    where is the normalized histogram:

    (10)

    and represents the desired histogram from the input imagehistogram. Based on logarithm function, it can be expressed as

    (11)

    where is the maximum value of input image histogramand is the parameter to control the intensity. According to theempirical law [14], is decided as 5.In order to employ the power constraint and image contrast

    enhancement term, the total dissipated power is employed as thevector form:

    (12)

    where is the vector as . On thepurpose of power saving and contrast enhancement, the PCCEapplies the convex optimization function to minimize thepower consumption and to improve the image contrast quality.To make the optimization function to be more precise, theeuclidean distance is used in the convex optimized problemand is formulated below:

    (13)

    where represents the control parameter between the powerconstraint and contrast enhancement. To balance contrastenhanced and power-saving, the generalized Lagrangian costfunction is then employed.

    B. HS Method [16]

    The basic concept of the HS algorithm is to shrink the inputimage histogram and to keep the image details. In order toachieve the purpose, it uses the following definition of imageentropy preservation:

    (14)

    where represents the gray-level and is the probabilityof the normalized image histogram. According to the defini-tion of logarithm function, the summation equation is subjectto . Suppose that and are the input image andoutput image information. The input image entropy and outputimage entropy are formulated as

    (15)

    (16)

    where represents the probability of of the input imageis the probability of of the output image .Based on the entropy preservation definition, must

    equal to the . In other words, the histogram of input

    image must shift accurately, i.e., prevent any two bins of theimage histogram from combining.

    C. DiscussionThe significance of the work [14] lies in the fact that it is

    the first approach to enhance the image contrast and reducethe power consumption for OLED displays in the meantime.Specifically, the PCCE method proposed in [14] formulated apower consumption model with an objective function, whichconsists of the histogram equalizing term and power term. Thesimulation results of PCCE algorithm show that it provideshigh image contrast and good perceptual quality while reducingpower consumption. The average power-reduction ratio isapproximately 36% for the test images. However, the convexoptimization equation uses a parameter that needs to bemanually set. Hence, this method is not suitable for all of theinput images/videos within the automatic process. On the otherhand, in [16], a novel HS method was proposed to decreasethe empty bins of input image histogram before using imagecontrast enhancement technique. In other words, it reducesthe power consumption before enhancing the image qualityand prevents the distortion of image histogram after powerreduction. Thus, the HS approach can be considered as aneffective pre-processing scheme for OLED displays. However,the contrast of the image cannot be enhanced simultaneously.Therefore, it is paramount to design a novel HE-based algo-rithm that can accurately balance the power consumption andcontrast enhancement automatically.

    III. PROPOSED SCHEMEIn this section, we will present the proposed novel algorithm

    that can reduce the power consumption and enhance the con-trast of the image concurrently. Therefore, a balance betweenthe power consumption and image quality can be achieved usingthe proposed approach.In order to reduce power consumption and enhance image

    quality at a emissive display which does not have the backlightmodule. To reorganize the image pixel value, an image pre-pro-cessing stage is employed. The proposed algorithm named aspower-saving histogram adjustment (PSHA).

    A. PSHA AlgorithmThe OLED has self-emissive properties so that the display

    luminance is dependent on the image pixel intensity, that is,the more low-intensity pixels the less luminance. Therefore, thePSHA algorithm could decrease the luminance of the imagepixel while at the same time keeping the details.Let be the input image and be the pixel index. In the

    beginning, it equally shift each pixel value. The shifted inputimage is expressed as

    (17)

    By using the above calculation, the minimum luminance levelof the image becomes zero to slightly reduce the display powerwhile preserving the original contrast. It is noted here thatdifferent images will have different minimum luminance level

    .

  • JAN et al.: POWER-SAVING HISTOGRAM ADJUSTMENT ALGORITHM FOR OLED-ORIENTED CONTRAST ENHANCEMENT 371

    The (3) is employed for to generate the corresponding PDFdenoted by . After performing PDF generation, it generate thedetection table to label the empty bins. Let be the levelindex. The pseudo-code of this process is expressed as1:2: for to do3: if then4:5:6: end if7: end for8:

    When the detection table is constructed, the length betweenthe non-empty bins can compute by1: for to do2:3: end for

    When the detection table is constructed, the length betweenthe non-empty bins can compute by

    (18)

    for the range of c from 1 to , where c denotes the graylevel. Thus, the average length is then computed as follows:

    (19)

    Furthermore, with the utilization of , and, the transformation function can be generated by the

    following process:

    (20)

    Based on this transformation function, the modified input imageis then calculated by

    (21)

    The major advantages of the proposed PSHA algorithm aredescribed below:1) Display power saving: it can normalize the luminance

    probability distribution, while the display power is not beincreased.

    2) Entropy consistent: it only modifies the image histogramand prevents the combination of any two bins, so that en-tropy consistent is achieved to preserve the image details.

    3) Low complexity: the time complexity of the PSHA algo-rithm depends on the number of luminance levels that ismuch lower than the image size. Hence, it is appropriate toreal-time applications.

    B. Adaptive PCCE Algorithm

    Through the proposed PSHA algorithm, the luminance ofthe modified image become lower than the original one, whilethe length of the successive empty bins is normalized to keepthe contrast. However, the input image still needs to furtherreduce the display power in the OLED panel. Based on theproposed PSHA algorithm, it combines the proposed adaptive

    PCCE (APCCE) algorithm that balances the contrast enhance-ment and power reduction effects. Compared to the empiricalsetting of the PCCE, an improved computation method is em-ployed in this paper to generate the adaptive parameter of (13).It has been shown in [14] that the power-saving can be gained

    by decreasing the grey level value, which represents the incre-ment of the control parameter . Considering the characteris-tics of the PSHA, the luminance of the modified image would belower than the luminance of the shifted image in the most cases.Therefore, a mean value of the modified and shifted image iscomputed to serve as the estimation term. The adaptive (de-noted by ) is calculated below:

    (22)

    where the and the represent the averageof the modified image and the shifted image respectively. Theconvex optimized problem in (13) can be reformulated by

    (23)

    After being computed by the (22), the parameter is thendependent on the automatic change of the luminance corre-sponding to the influence of power modified by the PSHAalgorithm. As a result, power consumption and image contrastcan be jointly considered in the proposed algorithm.

    IV. EXPERIMENTAL RESULTS

    In this section, we present experimental results including soft-ware simulations and real estimations of an OLED panel toverify that our proposed algorithm can achieve power-savingand contrast enhancement simultaneously. At first, simulationresults of image quality and power consumption will be illu-trated. In the following, we use a commercial OLED panel todemonstrate the current value (power consumed) and imagequality.

    A. Simulations of Image Quality and Power Consumption

    First of all, we evaluate the effects of the proposed algorithmand the approaches of [14] and [16] on six test images enti-tled test1–test6, respectively. Morover, the images test1–test3have resolution of 512 512 pixels and the images test4-test6have resolution of 768 512 pixels. In simulations, we trans-form the RGB color space of the input image to the YUV colorspace and only process the luminance component (the Y-com-ponent) using each algorithm. In all experiments, is set to 2.2as describe in [19]. Fig. 2 shows the enhanced images by dif-ferent methods. Compared with the proposed method, the othermethods need to tune the settings of the parameter for pro-cessing various images. As shown in [14], if is increased, thepower consumption of displaying the image would be reduced,but the image brightness is also degraded.Furthermore, to accurately measure the processing effect,

    we also perform the quantitative measurements to each outputimage. According to [7], we use the traditional EME metric tomeasure the contrast of the image. Let be the total numberof the segmented blocks in the image. Before computing the

  • 372 JOURNAL OF DISPLAY TECHNOLOGY, VOL. 12, NO. 4, APRIL 2016

    Fig. 2. Enhanced results of the test1–test6 images for (a) original images, (b) the proposed results, (c)–(e) PCCE [14] with respectively, and(f)–(h) PCCE with HS pre-processing [16] with respectively.

    Fig. 3. Quantitative contrast measurement (EME) for the enhanced test1–test6images and comparisons with PCCE [14] and PCCE with HS pre-processing[16].

    EME, the image must be segmented into equivalent blocksdenoted by . The EME metric is formulated below:

    (24)

    Note that the higher EME value, the high contrast of the testimage is. In addition to the measurement of the contrast, we alsorefer to TDP computation to simulate the power consumption ofdisplaying each image in the OLED panel. Figs. 3–4 show theEME and TDP values of the enhanced test images.

    Fig. 4. Power consumption measurement (TDP) for the enhanced test1–test6images and comparisons with PCCE [14] and PCCE with HS pre-processing[16].

    It can be observed that the proposed method does not needthe manual parameter setting to generate the suitable results thatalmost approximate the highest EME as well as the lowest TDPfor various images.

    B. Measurement of Power Consumption on an OLED PanelIn order to verify the performance of our proposed ap-

    proach, extended experiments have been conducted basedon a real OLED panel. We use a Samsung panel of modelAMS369FG06-0 3.7 with visual WVGA 480 800, 16 Mcolor, and 3.3 V supply voltage. The test bench setup of thisOLED panel is shown in Fig. 5. Moreover, for fitting the size of

  • JAN et al.: POWER-SAVING HISTOGRAM ADJUSTMENT ALGORITHM FOR OLED-ORIENTED CONTRAST ENHANCEMENT 373

    Fig. 5. Test bench setup for practical measurement: (a) original image; (b) PCCE [14] with ; (c) PCCE with HS pre-processing [16] and ; and(d) the proposed method.

    Fig. 6. Experiments on OLED panels using: (a) PCCE [14]; (b) PCCE with HS pre-processing [16]; and (c) the proposed method.

    TABLE IPOWER CONSUMPTION REPORT OF PCCE [14], PCCE WITH HS PRE-PROCESSING [16], AND THE PROPOSED METHOD

    OLED panel, we consistently re-set the test image resolution to480 800 and set the bit depth to 32-bit. In particular, we mea-sure the OLED panel current to illustrate power reduction forall test images including the proposed image, original image,PCCE image, and HS-based PCCE image. Table I presents thecurrent of OLED panel for various images, where it can beseen that the HS-based PCCE image results in lower currentthan that of the proposed image in some situations. However,the proposed method has better human perception than theHS-based PCCE method, as shown in Fig. 6. In addition, powerreduction rate can be derived from Table I. For example, the

    current for the original test image 1 is 5.68 mA and taht for theproposed method is 3.07 mA. This shows that that the proposedmethod can reduce 45.95% power in test image 1. As also canbe derived from Table I, the HS-based PCCE method can leadto an average power reduction of 63.24% when is equal to 1,while the proposed method achieves a comparable reduction of60.82%.

    C. Computational ComplexityWe analyze the computational complexity for the proposed

    method, [14], and [16] assuming an image with resolution of

  • 374 JOURNAL OF DISPLAY TECHNOLOGY, VOL. 12, NO. 4, APRIL 2016

    TABLE IICOMPARISON OF OPERATION FOR PCCE [14], PCCE WITH HS [16], AND THE PROPOSED METHOD

    . For the proposed method, the complexity is O(L), whereL is the grey level and is only related to the number of bits perpixel. For example, for a 8-bit pixel, L is equal to 256. In [16],there is only one subtraction operator to shrink the histogramand thus, for the image of pixels, the complexity belongsto O (MN). In [14], it uses logarithm function to modify everypixel which requires 3multiplications, 2 additions, 2 logarithms,and 1 division for each pixel. Therefore, the complexity of thePCCE belongs to O(8MN). It notes here that each method needsto solve the optimization problem. The complexity of it belongsto O(Ł ). As can be seen, the time complexity of the proposedmethod is the lowest.We further take an example to illustrate thenumerical results of the computation complexity. Consideringan image with resolution of 480 800 8-bit pixels, the numberof multiplication, addition, subtraction, division, and logarithmsfor each approach is summarized in Table II. As can be seen,the number of computations for the proposed approach is thelowest. In order to further compare the timing complexity, allalgorithms are implemented in Matlab R2011b and executed onan platform with Intel core i7 3.4 GHz processor with a 4 GBRAM. As can be seen in Table II, the proposed algorithm takes0.29 s to process an image.

    V. CONCLUSION

    We have proposed the PSHA algorithm to be the pre-pro-cessing of the contrast enhancement algorithms for OLEDpanels. We also have presented the embodiment that modifiedthe PCCE algorithm to adaptively generate the parameteraccording to the image characteristic. Compared to the priorwork, the proposed PSHA algorithm effectively balances theeffect between contrast enhancement and power reduction. Inour future works, the statistical data of the transformation curvewill be analyzed to construct Look Up Table (LUT) for thereal-time hardware implementation. Moreover, the approachmentioned in this paper is only applied to static images. One ofthe future directions would be investigating extended methodsthat are applied to video sequences.

    REFERENCES[1] J.-Y. Huang, P.-F. Kao, and Y.-S. Chen, “A set of image processing

    algorithms for computer-aided diagnosis in nuclear medicine wholebody bone scan images,” IEEE Trans. Nucl. Sci., vol. 54, no. 3, pp.514–522, Jun. 2007.

    [2] H.-Y. Cheng, C.-C. Weng, and Y.-Y. Chen, “Vehicle detection in aerialsurveillance using dynamic Bayesian networks,” IEEE Trans. ImageProcess., vol. 21, no. 4, pp. 2152–2159, Apr. 2012.

    [3] P.-H. Lee, S.-W. Wu, and Y.-P. Hung, “Illumination compensationusing oriented local histogram equalization and its application toface recognition,” IEEE Trans. Image Process., vol. 21, no. 9, pp.4280–4289, Sep. 2012.

    [4] J.-M. Guo, Y.-F. Liu, and J.-D. Lee, “License plate localization andcharacter segmentation with feedback self-learning and hybrid-bina-rization techniques,” IEEE Trans. Veh. Technol., vol. 57, no. 3, pp.1417–1424, May 2008.

    [5] S.-R. P. Silva et al., “Nanoengineering of material for field emissiondisplay technologies,” IET Circ. Devices Syst., vol. 151, no. 5, pp.489–496, Oct. 2004.

    [6] P.-S. Tsai, C.-K. Liang, T.-H. Huang, and H.-H. Chen, “Image en-hancement for backlight-scaled TFT-LCD displays,” IEEE Trans. Cir-cuits Syst. Video Technol., vol. 19, no. 4, pp. 574–583, Apr. 2009.

    [7] T. Celik and T. Tjahjadi, “Contextual and variational contrast enhance-ment,” IEEE Trans. Image Process., vol. 20, no. 12, pp. 3431–3441,Dec. 2011.

    [8] T. Celik and T. Tjahjadi, “Automatic image equalization and contrastenhancement using Gaussian mixture modeling,” IEEE Trans. ImageProcess., vol. 21, no. 1, pp. 145–156, Jan. 2012.

    [9] O. Ghita, D. E. Ilea, and P. F. Whelan, “Texture enhanced histogramequalization using TV- image decomposition,” IEEE Trans. ImageProcess., vol. 22, no. 8, pp. 3133–3144, Aug. 2013.

    [10] S.-J. Kang and Y.-H. Kim, “Multi-histogram-based backlight dimmingfor low power liquid crystal displays,” J. Display Technol., vol. 7, no.10, pp. 544–549, Oct. 2011.

    [11] Y.-K. Lai, Y.-F. Lai, and P.-Y. Chen, “Content-based LCD backlightpower reduction with image contrast enhancement using histogramanalysis,” J. Display Technol., vol. 7, no. 10, pp. 550–555, Oct. 2011.

    [12] S.-I. Cho, S.-J. Kang, and Y.-H. Kim, “Image quality-aware backlightdimming with color and detail enhancement techniques,” J. DisplayTechnol., vol. 9, no. 2, pp. 112–121, Feb. 2013.

    [13] Y. Li, P. Chu, J. Liu, and S. Du, “A novel partitioned light guide back-light LCD for mobile devices and local dimming method with nonuni-form backlight compensation,” J. Display Technol., vol. 10, no. 4, pp.321–328, Apr. 2014.

    [14] C. Lee, C. Lee, Y.-Y. Lee, and C.-S. Kim, “Power-constrained contrastenhancement for emissive displays based on histogram equalization,”IEEE Trans. Image Process., vol. 21, no. 1, pp. 80–93, Jan. 2012.

    [15] S.-C. Huang, F.-C. Cheng, and Y.-S. Chiu, “Efficient contrast enhance-ment using adaptive gamma correction with weighting distribution,”IEEE Trans. Image Process., vol. 22, no. 3, pp. 1032–1041, Mar. 2013.

    [16] Y.-T. Peng, F.-C. Cheng, L.-M. Jan, and S.-J. Ruan, “Histogramshrinking for power-saving contrast enhancement,” in Proc. IEEE Int.Conf. Image Process., Sep. 2013, pp. 891–894.

    [17] C. Lee, C. Lee, and C.-S. Kim, “Contrast enhancement based on lay-ered difference representation of 2D histograms,” IEEE Trans. ImageProcess., vol. 22, no. 12, pp. 5372–5384, Dec. 2013.

    [18] H. Xu, G. Zhai, X. Wu, and X. Yang, “Generalized equalization modelfor image enhancement,” IEEE Trans. Multimedia, vol. 16, no. 1, pp.68–82, Jan. 2014.

    [19] M. Dong, Y.-S. K. Choi, and L. Zhong, “Power modeling of graphicaluser interfaces on OLED displays,” in Proc. Des. Autom. Conf., 2009,pp. 652–657.

    Li-Ming Jan received the B.S. degree in electricalengineering from Tung Hai University in 2012and the M.S. degree in electronic and computerengineering from National Taiwan University ofScience and Technology in 2014.He is currently with the Mstar Co., Ltd. where he

    is a Senior Engineer. His research interests includethe digital image processing, low power design andcontrast enhancement for OLED and LCD display.

  • JAN et al.: POWER-SAVING HISTOGRAM ADJUSTMENT ALGORITHM FOR OLED-ORIENTED CONTRAST ENHANCEMENT 375

    Fan-Chieh Cheng received the Ph.D. degree inelectronic and computer engineering from NationalTaiwan University of Science and Technology,Taipei City, Taiwan.He has published more than 30 international

    journal and conference papers related to image pro-cessing. His current research interests include picturequality improvement for video surveillance, digitalcamera, smart phone, smart TV, and other modernmultimedia systems, such as contrast enhancement,noise reduction, haze removal, and content-adaptive

    brightness control.

    Chia-Hua Chang received the B.S. degree from theDepartment of Electrical Engineering, Feng ChiaUniversity, in 2014, and is currently working towardthe M.S. degree at the National Taiwan Universityof Science and Technology, Taiwan.His research interests include digital image

    processing, low-power design, and contrast enhance-ment for OLED display.

    Shanq-Jang Ruan received the B.S. degree incomputer science and information engineering fromTamkang University, Taiwan in 1995, the M.S.degree in computer science and information engi-neering in 1997, and the Ph.D. degree in electricalengineering from National Taiwan University in2002.He served as an Electronic Officer in R.O.C.

    Air Force during July 1997 and May 1999. FromSeptember 2001 to May 2002, he served as a Soft-ware Engineer with Avant! Corporation. From June

    2002 to November 2003, he was with Synopsys as a Senior Software Engineer.His research interests include power efficient circuits and systems, imageprocessing, and electronic design automation. He was an Assistant Professorwith the Department of Electronic Engineering, National Taiwan Universityof Science and Technology, from 2003 to 2007. He is currently a Professorwith the Department of Electronic Engineering, National Taiwan University ofScience and Technology.

    Chung-An Shen (M’10) received the B.Sc. degreefrom the National Taiwan University of Science andTechnology (Taiwan Tech), Taipei, Taiwan, in 2000,the M.Sc. degree from the Ohio State University,Columbus, OH, USA, in 2003, and the Ph.D. degreefrom the University of California at Irvine, Irvine,CA, USA, in 2012.He joined the Department of Electronic and

    Computer Engineering, Taiwan Tech, in 2012, wherehe is currently an Assistant Professor. His currentresearch interests include low-power digital circuits,

    and signal processing architectures for wireless communication and multimediasystems.

  • 本文献由“学霸图书馆-文献云下载”收集自网络,仅供学习交流使用。

    学霸图书馆(www.xuebalib.com)是一个“整合众多图书馆数据库资源,

    提供一站式文献检索和下载服务”的24 小时在线不限IP

    图书馆。

    图书馆致力于便利、促进学习与科研,提供最强文献下载服务。

    图书馆导航:

    图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具

    http://www.xuebalib.com/cloud/http://www.xuebalib.com/http://www.xuebalib.com/cloud/http://www.xuebalib.com/http://www.xuebalib.com/vip.htmlhttp://www.xuebalib.com/db.phphttp://www.xuebalib.com/zixun/2014-08-15/44.htmlhttp://www.xuebalib.com/