Home > Engineering essays > Enhancement For Oled Displays Based On Histogram Equalization

Essay: Enhancement For Oled Displays Based On Histogram Equalization

Essay details and download:

  • Subject area(s): Engineering essays
  • Reading time: 12 minutes
  • Price: Free download
  • Published: 7 June 2012*
  • File format: Text
  • Words: 2,330 (approx)
  • Number of pages: 10 (approx)

Text preview of this essay:

This page of the essay has 2,330 words. Download the full version above.

Enhancement For Oled Displays Based On Histogram Equalization

ABSTRACT-
Power constrained contrast enhancement for organic light emitting diode ( OLED) displays based on histogram equalization is proposed to enhance an image by constraining power in emissive displays like LCDs , OLEDS etc , and the same can also be extended on to videos sequences. Initially an image is considered and the over stretching artifacts, sharpness and other effects are enhanced using histogram equalization techniques. Then a power consumption model is constructed and incorporated with the histogram equalized output formulating a function that consists of both the histogram equalization term and the power term. Then using Convex optimization theory by adjusting the different values of the formulated function , power saving and contrast enhancement are achieved simultaneously showing the simulation results using matlab
‘ INTRODUCTION :
An image is a group of pixels arranged in the form of matrix of any order say m*n (m is the number of rows and n is the number of columns). Using the matrix size, pixel intensities and other features we perform different image operations depending on our requirement .using the latest imaging technology we can process any type of digital image and remove all unnecessary distortions .During this image processing in order to achieve higher contrast and perceive the image clearly more power is consumed by the display panels . This is an issue with most of the consumer portable electronic gadgets like mobile phones ,tablets computers , GPS (global positioning system) , as they cannot be operated for long time with greater contrast .it is a known fact that for LCD based gadgets 90% of the power consumption is attributed to its backlight. Accordingly saving the backlight is considered the best solution to save power consumption ,but when a user tries to constrain the power we obtain lower quality and contrast images. Great effort have been made to increase the sharpness, contrast ,color , accuracy etc. and simultaneously reduce power consumption in the displays and achieve a better perception for the viewer .For this reason it is essential to develop an image processing algorithm that is capable of saving power in the display panel and also increase the contrast at the same time. Present modern display panels are generally divided into two , namely emissive and non-emissive display panels . examples for emissive displays are cathode ray tubes (CRTs) ,plasma display panels (PDPs) ,organic light emitting displays (OLEDs),field emissive displays (FEDs) whereas the thin film transistor liquid crystal displays (TFT-LCDs) are examples of non-emissive displays. The main feature of emissive displays that distinguish itself from non-emissive displays is that the emissive displays do not require any external light sources unlike the non-emissive displays and they can achieve higher contrast with less power consumption.An emissive display can turn off individual pixels independently to express darkness there by increasing contrast ratio. The power consumption of each pixel is proportional to its intensity level , hence proving that an emissive display consumes less power than a non-emissive display in which the backlight must be turned on regardless of pixel intensities.

There have been numerous approaches to improve the contrast and reduce the power consumption in emissive displays .these approaches mainly focus on reducing the backlight while preserving the sense of vision and quality even though pixel intensities are increased to compensate the reduce in contrast.
In this paper a power constraint contrast enhancement algorithm for emissive displays (like OLEDs) based on histogram equalization is proposed. Initially a histogram equalization model is used to eliminate the overstretching artifacts in the test image and then the same output is modified using a modified histogram model to reduce large histogram values to alleviate the contrast overstretching . Then we construct a power consumption model formulating a function which consists of the histogram equalization term and the power term. To operate with the created function we employ a process called convex optimization technique. By minimizing the function based on convex optimization the algorithm achieves contrast enhancement and power consumption simultaneously. The same algorithm can also be extended to video sequences which are shown in the simulation results at the end .
Figure 1a . Histogram equalization applied to a low contrast image
‘ Convex optimization :
Sparse representation of a given image is expressed as a linear combination using a few non-zero coefficients in an appropriate basis-function.The convex optimization method achieves state-of-the-art performance in compressed sensing reconstruction and digital image de-blurring applications.
‘ Histogram Equalization Techniques :
Histogram equalization technique is an approach to enhance low contrast images by making a histogram of the light intensities of all pixels as uniform as possible increasing the overall dynamic range . This can be used on any part of an image or the complete image without introducing any new intensities in an existing image. Here is a review of various histogram equalization (HE) techniques.
‘ Histogram Equalization (HE): This method is the simplest model to enhance the contrast of an image. Let F be a given image with matrix of order Mr * Mc with all the pixel intensity values ranging from 0 to (L-1) where L represents the possible number of intensity values which is probably 255 . now P be the normalized histogram of F .
Therefore Pn =Number of pixels with intensity n/Total number of pixels , where n= 0,1,’..(N-1)
The histogram equalized image G is given by
Gi,j = floor ((L-1) ‘.(a1)
where floor () tends to be the nearest possible integer
The pixel intensities K of F is given by the equation T(k) = floor ((L-1)) . the idea of transformation is based on considering the intensities of F and G as the continuous random variables X,Y on 0 to (L-1) with Y given by equation.
Y = T(X) = (L-1) , ‘.a(2)
Where Px is the probability density function of F and T is the cumulative distributive function of (X *(L-1)) Let us assume that T is differentiable and invertible , it can now be defined as T(X) be uniformly distributed on (0 to (L-1)) ,namely PY(y) = 1/(L-1)
= probability 0 ‘ Y ‘ y.

Figure1originalimage Figure 2 originalhistogram

Figure3 transformed image , figure 4 transformed histogram

‘ Histogram modification (HM) :Simple histogram equalization technique has several disadvantages , for instance when the bin values are in large numbers the histogram transformation function takes an extreme steep slope leading to image alteration in an undesired fashion .also in HE image intensity is increased to a great extent and simultaneously some undesired noise arises degrading the resulting image . To overcome these drawbacks few more techniques are introduced ,one of them is Histogram Modification technique. In this method its possible for us to control and adjust the level of histogram equalization to enhance an image as with greater accuracy with less distortion unlike in the simplified histogram equalization process .The modified histogram is the average of the input histogram and the equalized uniform histogram given by the equation

‘. (b1)

The histogram of the input image for the modified histogram is given by
Km = 1/(1+Lm)’.(b2)

‘ Logarithmic histogram modification : in this project for simplicity purpose in order to reduce large pixel intensity values we develop a histogram modification method using a logarithmic function which is monotonically increasing this scheme is called as Logarithmic Histogram Modification (LHM).The obtained logarithm function is used to convert the input Histogram Hi into a modified Histogram Hm by using the following equation

Mk ”(b3)
Where, Hmax is the maximum value of input histogram H, and ,m is the coefficient that controls the standard of histogram modification.
From the above equation if m increases the term (log(Hk*Hmax*[10-m] minimizes and as m decreases the term maximizes hence larger m value makes Mk proportional to Hk.
As Log (1+x) ~ x for a smaller x leading to weakly modified histogram and , when m is small (log(Hk*Hmax*[10-m] ~ Log(Hmax*10-m) Leading to a strongly modified histogram.
In other words we can express Hm =F(H) , where Hm = [Hm0 ,Hm1′.Hm(L-1)]t. The transformation function x = [ x0,x1,’.x(L-1)]t is expressed as [D*x] = Hm’
The figure below shows a sample to illustrate LHM :

‘Image source from ref [5]’
Figure 2a ) histogram of the input test image as in figure c. figure b) histogram of the normalized image ,figure c)original input test image ,figure (d ) output for histogram equalized image ,
Figure (e) output image for m=2, figure (f) output image for m=5, figure (g) output image for m=7.
In the above illustration we can observe different outputs for different values of m which is also shown for similar images in the simulation results.
‘ Power Constrained Contrast Enhancement (PCCE ):
In this project we use the above described techniques to develop a PCCE algorithm and the overview of the algorithm is shown in the block diagram ( figure 4.1).

Figure 4.a Block diagram of the proposed PCCE Algorithm ‘image source from ref 5’

As shown in the above figure (4.a) , we first concider a test image and acquire the histogram information H of the same and then applying Logarithm histogram modification technique on the histogram H obtained initially and obtain a modified histogram Hm of the input image then without constraining power as per the expression [ D*x ] = Hm’ we can calculate the transformation function x.
Using that function we design an objective function in terms of a variable Y using expression Y =[ D*x], now using the Convex optimization concept we calculate a minimul Y that helps in optimizing the objective function finally the transformation function x is calculated using Y using x = (D-1) *Y and then using the value of x , the input test image is converted obtaining the required output.
‘ Power model for emissive displays:

for the images to display we need to design a power consumption model .According to the experimental results it is proved that the power P of a single color pixel is given by equation
P = Wo + Wr*(Rg) +Wg*(Gg) +Wb*(Bg) ,”(4.1)
where R , G , & B are the red , blue and green components of the pixel and ‘ g ‘ is the gamma correction color value , Wo gives the static power consumption ( independent of any pixel values) ,Wr,Wg,Wb are the weighing coefficients which expresses the different characteristics of the RGB pixel components .In general ‘ g ‘ has the value of 2.2 for optimal power consumption purpose .After transformation of the color values into the RGB

format we can obtain a liner relationship between power and the luminous intensities.
For the paper we need to alter the pixel values to save the power in the displays .we can calculate the total power dissipated in an image using the expression for an RGB image,
TPD = ‘.(4.2)
Where n is the numer of pixels in the image .Ri,Gi,Bi are the rgb color values at the (i^th ) pixel.
Generally the weight coefficients are inversely proportional to the subpixel efficiencies In an image and depends on the characteristics of the display panel of any kind. It’s a fact that blue pixels consume more power than the rest of the color pixel components (red and green) when displaying the same output due to low efficiency. Also in the similar way for a gray scale image we have the expression for total power dissipated given by:
TDP = ‘.(4.3)
Where Yi is the Gray level of the Ith pixel ,to explain we consider there are Hk pixels with gray level K in the test input image, let the gray levels of k pixels be Xk in the obtained output image by using the transformation function .hence we can rewrite the total dissipated power in vector notation as :
TDP = = ”..(4.4),
where (x) = (Xog ),(X1g) .. X(L-1)g ),and H is the vector histogram with Hk as the kth element.
Both the power models described in equation (4.1) and (4.2) are applicable to almost all kinds of emissive displays.
‘ Constrained optimization problems :

The main aim of the project is to increase the contrast of an image and simultaneously reducing the power consumption.In reality both the tasks are contradictory to achieve at the same time. And to achieve this task at the same time we apply the above described power disscipation models and histogram equalization. These two tasks are stated as a constrained optimization problem as explained below
Minimization of (||Dx ‘ M’||2) +A(Ht) (x)’.(4.5) . And making Xo = 0 , X(L-1) =L-1, Dx => 0′..(4.6)
Where ( || Dx ‘ M’ ||2 ) is the histogram equalization term and [ (Ht) (x) ] is the power term as in equation (3) and A is the controlling parameter that we can use to balance the equation.
. Moreover the power term is usually proportional to the average luminance value of input, and we can modify the variable’ B ‘as
B =A’.(4.7)
By minimizing these above two terms in equation (4.5) we can achieve the task of increasing the contrast of the image and simultaneously reducing the power consumption .
The three constraints as mentioned in equation (4.6) state that intensity levels of the image must be maintained without making any major changes in the image.ie if a display is capable of presenting L different pixel intensity levels , the output transformed must be in the range of [ 0 ‘(L-1) ] to exploit the full dynamic range . the equation Dx => 0 implies x to be monotonic which means xk >=x(k-1) for all vales of k. without this monotonic constraint function the yielded output transformation function may reverse the intensity ordering of pixels producing visually annoying artifacts in the output image.
‘ Power consumption contrast enhancement for video sequences :
We extend the same procedure as described above to video sequences which also gave the same output with reduced power and increase in contrast using the power control parameter ‘ B ‘ as given in equation (4.7) . we can apply different values of B and compare the video sequences , as we increase B the ,more power is saved but after a certain level if we further increase B the brightness might get distorted. To apply PCCE For a video sequence of frames we need to set a target power to be consumed and accordingly input power is controlled .
TPD out =K( TPDi ) ‘.(5.1)
Where k is the power reduction ratio , for example if k=1, no power is saved ,and when k <1 , the power consumption decreases and also darkens the output video sequence therefor we can say more power is saved with brighter video sequence . The power reduction ration ,k is given by
K = ‘.(5.2)
Where Y’ is the average grey level of an input image or video frame, r is the controlling parameter. Which means for a higher value of Y’ and brighter frame k becomes smaller to achieve power saving and vice versa.. as a summery , we take a vide sequence first then determine the target power to be consumed using equation (5.1) and (5.2) by setting value of B and controlling the power using the parameter r, larger r produces smaller k which inturn saves more power.

‘ EXPERIMENTAL RESULTS
We have performed the power constrained contrast enhancement algorithm on three different test images and evaluated the performance as obtained .
In the figure (6.1) obtained below the difference between a power constrained image and also a no power constrained (bet = 0) image is shown and it is clearly evident that as the value of ‘ beta’ increases the power consumption is reduced , given ‘ beta’ is the power term .we performed the PCCE algorithm taking different beta values from ‘ 0 ‘ , ‘ 0.5 ‘ , ‘ 2.84 ‘ and ‘ 3.0 ‘ . Finally the total dissipated power is noted for all values of beta.
In figure (6.21, 6.22, 6.23) we can see the histogram modified images with their corresponding histograms for different values of ‘ mu ‘ like ‘ 2 ‘ , ‘ 5 ‘ , ‘ 7 ‘ .where mu is the parameter that represents the level of histogram
Proving that a smaller value of mu yields stronger histogram and a larger value of mu yields a weak histogram .

 

 

2a
3a
4a
5a
6a

2b
3b
4b
5b
6b

2c
3c
4c
5c
6c

Figure 6.1 . 2a, 2b , 2c : original images , 3a,3b,3c : image output with no power constraint (bet = 0) and increased contrast , 4a , 4b , 4c : PCCE output with bet = 0.5 ,5a,5b,5c : PCCE output with bet = 2.8 ,
6a, 6b, 6c : PCCE output with bet = 3.0
Total Dissipated Power of Output image , ( TDPout ) = 1.7756( fig .3a) , 1.0746 (fig 4a) , 0.4072fig (5a), 0.4213 fig (6a )
Total Dissipated Power of Output image , ( TDPout ) = 1.6217 ( fig .3b) , 0.8704 (fig 4b) , 0.2981fig (5b), 0.3088 fig (6b )
Total Dissipated Power of Output image , ( TDPout ) = 3.6953 ( fig .3c) , 2.2418 (fig 4c) , 0.8308fig (5c), 0.8628 fig (6c )

7 8
9

7a
8a
9a

Figure 6.21 ,7: PCCE output at mu = 2 , 7a : Histogram analysis at mu = 2, : 8 : PCCE output at mu = 5 , 8a : Histogram analysis at mu = 5, : 9 : PCCE output at mu = 7 , 9a : Histogram analysis at mu = 7,

4

5
6

4a
5a
6 a

Figure 6.22 ,4 : PCCE output at mu = 2 , 4a : Histogram analysis at mu = 2, : 5 : PCCE output at mu = 5 , 5a : Histogram analysis at mu = 5, : 6 : PCCE output at mu = 7 , 6a : Histogram analysis at mu = 7,

1
2
3

1a
2a
3a
Figure 6.23 , 1 : PCCE output at mu = 2 , 1a : Histogram analysis at mu = 2, : 2 : PCCE output at mu = 5 , 2a : Histogram analysis at mu = 5, 3 : PCCE output at mu = 7 , 3a : Histogram analysis at mu = 7

‘ CONCLUSION :
A power constrained contrast enhancement ( PCCE ) algorithm capable of enhancing image contrast and simultaneously reducing power consumption is proposed in the project . firstly a power consumption model is formulated and implemented with histogram equalization , formulating an objective function consisting of both power term and the histogram equalization term . in order to control both the terms , there arise some convex optimization problems and found the solution by deriving an efficient algorithm to find the optimal transformation function. using LHM scheme In the paper the simulation results show that we can achieve power consumption significantly while maintaining satisfactory image quality . also the same algorithm is applied to a video sequence and the simulation results are satisfactory.

References
[1] R.C. Gonzalez and R.E.Woods , Digital Image Processing ,3rd edition , Upper Saddle River , NJ ,Prentice-Hall 2007.
[2] CHU woo Lee : chul Lee; Chang-Su kim,’Power constrained contrast enhancement for OLED display based on histogram equalization ‘ , Image Processing (ICIP) ,2010 17th IEEE International Conference sept 2010
[3] Seungjoon Yang ;Jae Hwan Oh, Yungfun pank ‘ contrast enhancement using histogram equalization with bin underflow and bin overflow’ Image Processing ,2003. ICIP 2003.Proceedings , 2003 International conference Sept 2003
[4] Gyu-Hee Park ; Hwa-Hyun Cho ; Myung-Ryul Choi , ‘A contrast enhancement method using dynamic range separate histogram equalization ‘ , consumer electronics , IEEE transactions , November 2008.
[5] Chulwoo Lee; Chul Lee; Young-Yoon Lee; Chang-Su Kim, "Power-Constrained Contrast Enhancement for Emissive Displays Based on Histogram Equalization," Image Processing, IEEE Transactions on , vol.21, no.1, pp.80,93, Jan. 2012

...(download the rest of the essay above)

About this essay:

If you use part of this page in your own work, you need to provide a citation, as follows:

Essay Sauce, Enhancement For Oled Displays Based On Histogram Equalization. Available from:<https://www.essaysauce.com/engineering-essays/power-constrained-contrast-enhancement/> [Accessed 29-03-24].

These Engineering essays have been submitted to us by students in order to help you with your studies.

* This essay may have been previously published on Essay.uk.com at an earlier date.