In the coming of the Internet, a batch of different web pages and multimedia applications have been developed. Developers use images in their design to do it more appealing to the end-users but utilizing legion images in the design of applications will increase download clip. This is where image compaction is utilised.
In the past 15 old ages, legion types of image compaction have been developed. It attracted important attending due to the increasing sum of information we are utilizing mundane. Image compaction package is widely used to hive away or convey information by extinguishing excess information and therefore, minimising physical infinite [ 18 ] .
Previous surveies have used different optimisation algorithm to accomplish image compaction. Ant settlement optimisation, familial algorithm and atom drove optimisation was used and has already proved its worth in the optimisation of image compaction.
In 2001, Geem et. al [ 5 ] developed an optimisation algorithm called harmoniousness hunt. Harmony Search Algorithm is a music-inspired algorithm that mimics the attack of instrumentalist in looking for harmoniousness in playing music. Although this optimisation algorithm is comparatively new, old surveies have shown its capablenesss in work outing optimisation jobs in different Fieldss like technology, transit 1and environmental systems.
In this paper, the research worker will utilize Harmony Search Algorithm in optimising lossy compaction image.
Image is a planar image that resembles the visual aspect of the topic. It is besides referred to as the similitude seen or produced from something or person [ 10 ] .
An image is composed of an array of Numberss, runing from 0 to 255, that represent light strengths at assorted points. These points are referred to as pel or image component. And these pels are what make up a raster information. The pels in the image contribute to its file size. Each pel uses 3 bytes to stand for the colour – primary colourss: ruddy, green, and blue. When conveying this image, the file size becomes a major factor. And for this ground, image compaction would be good [ 2 ] .
This survey focuses on digital image. Two major types of digital image are the vector and raster ( besides called Bitmap ) . Vector is made up of single objects that are derived by mathematical statements and belongingss sing its colour and fill. This type of image can hold its highest quality in any graduated table since it its declaration independent [ 21 ] . On the other manus, raster or electronic image is composed of pel that contains the information on the colour to expose. This type of image has fixed declaration and losingss image quality when resized or displayed in another graduated table [ 1 ] .
Taken from [ 19 ]
Figure 1.1: A Raster Image
Although vector has similar or even better capablenesss compared to raster, it has non been popular. One of the grounds is its inability to bring forth photo-realistic imagination. It does non decently picture the uninterrupted elusive tones of the exposure due to solid country of colour. However, with the latest promotion in image processing, vector tools are now able to bring forth bitmapped texture to the objects [ 2 ] .
Taken from [ 15 ]
Figure 1.2: Difference of Vector and Bitmap Images
The construct of image analysis is widely used in computing machine or machine vision. It is used to pull out measurings and information from digital image through image processing techniques. It derives important information with respects to the selected image characteristic such as countries, size distribution, length, etc. [ 2 ] . Image analysis has been used as the underlying construct of saloon codification reading and face acknowledgment.
RGB function and numeral values display are merely 2 of the tools that are chiefly used in image analysis. This has made important impact on analysing images that are used in different field such as medical specialty, microscopy, remote detection, etc. [ 11 ] .
1.1.2 Image Processing
A planar image is treated as the input of image processing and outputs a modified image or set of features specifying the image. It involves transmutation and techniques that were derived from signal processing. Basic transmutation in image processing includes expansion, size decrease and rotary motion [ 13 ] .
Image compaction has big part in the success of execution of computing machine vision, characteristic sensing and augmented world. It has besides part of in the field of medical specialty. Medical image processing and medical image processing are merely the few of them.
1.1.3 Evaluation of Image Quality
To measure the quality of an image compared to a certain image, some quantitative steps are used. Two of the most widely used tools are the Mean Squared Error ( MSE ) and Peak Signal-to-Noise Ratio ( PSNR ) .
220.127.116.11 Mean Squared Error ( MSE )
Mean Squared Error is a tool used to quantify the mistake between two images. It can besides be defined as the norm of the square of the mistake. Figure 1.3 shows the expression in happening the MSE. Lower MSE value is better which means that there is less difference between two images [ 2 ] .
Figure 1.3: Formula for MSE
18.104.22.168 Peak Signal-to-Noise Ratio ( PSNR )
Peak Signal-to-Noise Ratio estimates the quality of the tight image compared to the original image. This is measured in dBs ( dubnium ) . Figure 1.4 the expression for work outing PSNR. A high PSNR value means that the quality of the tight image is less debauched when compared to the original image [ 2 ] [ 11 ] .
Figure 1.4: Formula for PSNR
1.2 Image Compaction
Image compaction is really good in image transmittal over the Internet and downloading from Web pages. Its chief intent is to minimise the size in footings of bytes of a in writing file without impacting the quality to an unacceptable degree. Decrease of file size of an image leads to larger sum of informations stored in same sum of memory infinite and lesser sum of clip required to direct an image [ 12 ] . Nowadays, a batch of different compaction applications are available for use but the resulting image is less than optimal.
There are two image compaction types. These are lossless and lossy compaction. These two types are differentiated based on the recovery of the image when decompressed.
1.2.1 Lossless Image Compression
Loseless compaction is a type of informations compaction technique wherein no information is lost and it retains the full information needed to retrace its original image. Since no information is lost, its compaction is someway delimited and can compact the original image merely approximately 50 % of its original file size [ 2 ] [ 16 ] .
Artworks Interchange File ( GIF ) and Portable Network Graphics ( PNG ) are merely two of the widely used loseless image compaction. Both of these formats support transparence but GIF can back up life. Because of its broad support and portability, its use on the web has increased.
1.2.2 Lossy Image Compression
From the name itself, lossy compaction is another type of informations compaction wherein some of the information of the original image is for good removed. Excess information is eliminated by which users may non detect it [ 2 ] . Since some informations are for good eliminated during compaction, there is debasement of the ocular quality of the image.
One of the most widely used lossy compactions is the Joint Photographic Experts Group ( JPEG ) . This is normally used in hive awaying and directing images on the Web. Its compaction is based on the ground of the inability of the human oculus to distinct some delicate colour and high frequence brightness fluctuations of an image.
1.2.3 Compression Ratio
Compression ratio is a term in computing machine scientific discipline that is used to mensurate the decrease in footings of informations representation size produced by the informations compaction algorithm. It shows the compaction power of an algorithm and files size the most normally used parametric quantity in mensurating the compaction ratio. The expression for compaction ratio is defined as:
1.3 Color Space
Color infinite is an abstract theoretical account used in the method of stipulating, making and visualising colour. It has three chief properties ; impregnation, brightness and chromaticity. A colour can be identified in a colour infinite by these properties bespeaking its place within the colour infinite and it is normally a transition or combination of colourss. There are six different colour infinites at the present that are used in image and picture processing ; RGB ( Red Green Blue ) , CMYK ( Cyan Magenta Yellow Black ) , HSL ( Hue Saturation and Lightness ) , YUV, YCbCr, YPbPr. Compaction applications have used RGB, YUV and YCbCr among the six bing colour infinites [ 2 ] .
RGB stands for ruddy, green blue constituents which are the primary colourss of visible radiation. This colour infinite is widely used both in image processing and computing machine artworks. Basically composed of three colourss, all other colourss can be achieved by uniting these three colourss. In the field of image processing, RGB represents the colourss that each pel in the image shows. The scope value of RGB is 0 to 255 [ 2 ] .
RGB is non an efficient colour infinite when function colourss in rendering standard picture show. And this is the ground why it has to be converted. Figure 1.6 shows the equations on how to change over RGB to YUV which is another colour infinite [ 2 ] .
Taken from [ 2 ]
Figure 1.5: Image split into its RGB constituents
RGB to YUV Conversion
Y = ( 0.257 * R ) + ( 0.504 * G ) + ( 0.098 * B ) + 16
U = – ( 0.148 * R ) – ( 0.291 * G ) + ( 0.439 * B ) + 128
V = ( 0.439 * R ) – ( 0.368 * G ) – ( 0.071 * B ) + 128
Figure 1.6: Formula for RGB to YUV Conversion
YUV is a colour infinite that takes consideration on human ocular perceptual experience. Ocular perceptual experience focuses on the sensitiveness of the human oculus to brightness and colour differences. This type of colour infinite uses luminosity ( Y ) and chrominance colour constituents ( U and V ) of a colour. The scope value of YUV is 16 to 235 [ 2 ] .
YUV to RGB Conversion
B = 1.164 ( Y – 16 ) + 2.018 ( U – 128 )
G = 1.164 ( Y – 16 ) – 0.813 ( V – 128 ) – 0.391 ( U – 128 )
R = 1.164 ( Y – 16 ) + 1.596 ( V – 128 )
Figure 1.7: Formula for YUV to RGB Conversion
Taken from [ 2 ]
Figure 1.8: Image split into its luminosity and chrominance constituents
1.4 Metaheuristic algorithms
Many of the real-world jobs are focused on seeking for productiveness and necessitate optimisation. And for these jobs, the find of optimal solution could non ever be guaranteed. Alternatively of utilizing exact methods, a more suitable attack is heuristics. It uses approximative methods utilizing iterative test and mistake attack in work outing for the best solution. Most of the heuristics are nature-inspired and a latest development of this method is the metaheuristic [ 3 ] .
Metaheuristic autumn under a larger field of algorithm called Stochastic Optimization. This type of optimisation algorithm uses some degree of entropy to seek an optimum solution. It is used in work outing jobs where the hunt infinite for the solution is excessively big for beastly force method. The finding of a good solution is done by proving a solution and measuring its public presentation. Metaheuristic algorithms have two basic map degree Fahrenheit and g. Function degree Fahrenheit generate random solutions while map g evaluates the public presentation of a peculiar solution [ 5 ] [ 6 ] [ 17 ] [ 20 ] .
Different attack has been used in old surveies done on image compaction. Ant Colony Optimization and Genetic Algorithm, which are discrepancies of metaheuristic algorithms, are merely few of them that were used in old surveies. In this survey, a comparatively new discrepancy of metaheuristic called Harmony Search Algorithm will be used in the execution of image compaction.
1.5 Harmony Search Algorithm
Computer scientists have found a connexion between playing music and happening optimum solution. And this relationship has led to the creative activity of a new algorithm called Harmony Search. Harmony Search was foremost developed by Geem et Al. in 2001. Though this metaheuristic algorithm is comparatively new, its effectivity and efficiency has been shown in assorted applications. Since its creative activity in 2001, this metaheuristic algorithm has been applied to many applications which include design of H2O distribution webs, groundwater mold, energy-saving despatch, truss design, vehicle routing, map optimisation, technology optimisation and others [ 20 ] .
Taken from [ 20 ]
Figure 1.9: Set of Jazz Instruments
Harmony hunt is a music-based optimisation algorithm. The purpose of music to seek for perfect harmoniousness is what inspired the creative activity of this metaheuristic algorithm. Finding harmoniousness in music is correspondent to happening an optimum solution in an optimisation procedure. For illustration in an orchestra or a set of wind instrumentalists, each instrumentalist assigned to his ain instrument plays a note lending to the entire quality of the harmoniousness of the music produced [ 5 ] [ 20 ] .
Taken from [ 20 ]
Figure 1.10: Relationship of Music Improvisation and Optimization
When in hunt for the best harmoniousness, a musician implements one of the three methods possible to come up with possible optimum elements: ( 1 ) Playing from memory, ( 2 ) Pitch Adjustment and ( 3 ) Randomization [ 5 ] [ 6 ] [ 20 ] .
1.5.1 Harmony Search: Creating Good Music
The music assigned to each instrument, for illustration a wind set, will lend to the overall harmoniousness of the music being produced. For instrumentalists to play or improvize music it can utilize of combination of the three methods viz. : ( 1 ) Playing music based on his memory, ( 2 ) Playing music comparable to the music on his memory and ( 3 ) Creating music through random notes.
1.5.2 Harmony Search: Searching for Optimum Solution
Back in 2001, Geem et Al. has formalized the three elements of the freshly developed optimisation algorithm. The three matching constituent are ( 1 ) The usage of harmony memory, ( 2 ) Pitch Adjustment and ( 3 ) Randomization. Each of these elements plays a critical function in seeking for optimum solution in Harmony Search Algorithm [ 5 ] [ 6 ] [ 20 ] .
For instrumentalists to make a good music, he can see bing composing. In the instance of Harmony Search, utilizing harmony memory ensures that possible solutions are stored as elements in the new solution vector. Another manner a instrumentalist can play good music is by playing music relation to an bing composing. Harmony Search besides uses this construct called pitch accommodation mechanism. This is the 1 responsible for bring forthing solutions that are somewhat varied from the bing solutions. Flip accommodation is besides referred as the development mechanism of Harmony Search Algorithm. Randomization comes in to play as the 3rd method in HS. This ensures that the hunt for the solution is non limited in the local optima. It makes the solution set more diverse and it is referred to as the geographic expedition mechanism of Harmony Search Algorithm [ 5 ] [ 6 ] [ 20 ] .
Taken from [ 20 ]
Figure 1.11: Harmony Memory Illustration
After low-level formatting, the optimisation procedure starts and terminates until expiration status is reached. The optimisation in harmony hunt algorithm is done per determination variable. The value of each determination variable is decided with regard to harmony memory credence rate ( rhmcr ) on each base on balls. The rhmcr decides if the value of the ith variable will be taken from the values in the harmony memory [ 5 ] [ 6 ] [ 20 ] .
D + pitch accommodation
Taken from [ 20 ]
Figure 1.12: Generating New Solution from Harmony Memory
1.5.3 Harmony Search Algorithm Pseudo codification
Harmony Search Algorithm
Define nonsubjective map degree Fahrenheit ( x ) , ten = ( x1, x2… xd ) Thymine
Define harmoniousness memory accepting rate ( raccept )
Define pitch seting rate ( rpa ) and other parametric quantities
Generate Harmony Memory with random harmoniousnesss
While ( t & A ; lt ; max figure of loops )
While ( one & A ; lt ; = figure of variables )
If ( rand & amp ; lt ; raccept ) Choose a value for the variable I
If ( rand & amp ; lt ; rpa ) Adjust the value by adding certain sum
Else Choose a Random Value
Accept the new memory if better
Find current best solution
1.5.4 Harmony Search Algorithm Flowchart
2 Literature Review
Image compaction has been really good in informations transmittal and informations storage. It makes the file size of an image smaller which allows more informations to be stored in same sum of disc infinite and lessens the clip of an image to be downloaded. Different applications have been developed to accomplish image compaction but the end product image is less than the optimum.
Surveies sing image compaction or cryptography have used MSE ( Mean Squared Error ) and Peaks Signal-to-Noise Ratio ( PSNR ) to measure the quality of the image produced compared to the original image. Through this rating, research workers are able to see the efficiency and effectivity of the public presentation of their proposed method [ 2 ] .
In the paper of Donoho et. al [ 4 ] , they discussed the relevancy of harmonic analysis in informations compaction particularly those that focused on wavelet-based compaction. They indicated the challenges in accomplishing an optimum information compaction. Those are ( 1 ) obtaining accurate theoretical accounts for the existent information, ( 2 ) obtaining optimum representation of the theoretical account, and ( 3 ) quickly calculating the optimum representation. The paper cited two important developments in informations compaction particularly in image compaction through harmonic analysis ; ( 1 ) fast cosine transform for JPEG criterion and ( 2 ) fast ripple transform fro JPEG-200.
Merlo, et. Al presented a new method in image compaction by utilizing familial algorithm, a metaheuristic, as the underlying algorithm [ 4 ] . In this method, they considered a solution on the bunch job based on familial algorithm. It searches for the optimal solution by coincident consideration and use a set of possible solution. They defined their fittingness map that minimizes the upset on the elements they were telling. Through this, it would take the confusion of the fittingness map and the bunch method. They apply the bunch as the last measure of the process when the ordered representation is already obtained. Harmonizing to the consequences they obtained, familial algorithm was an effectual optimisation algorithm for constellating therefore ensuing to image compaction. However, their proposed method has shown restrictions when covering with larger images since longer computational clip is required.
In the survey of M. Mohamed Ismael et. al [ 14 ] , they were able to research the feasibleness of Particle Swarm Algorithm ( PSO ) in constellating based image compaction. The optimisation algorithm was used to better the truth of the anticipation maps in raising strategy. The lifting strategy was used for better designation of forms in the image and attempts to extinguish redundancy in an optimum mode. Using the standard information set in image processing, the proposed attack used in this survey has shown the feasibleness of PSO with promising consequences in image compaction.
Amiya Halder et. al [ 8 ] used the construct of block optimisation and byte pressure in their survey to accomplish image compaction. In the block optimisation stage of image compaction, they divided the image into 4×4 pixel block. To be able to cut down the figure of different colourss in the image, they get the norm of the pel values within the block and so replace all the pels in the block with the value derived. Byte compaction was used in this procedure to be able to stand for the colourss in the image utilizing fewer sums of bytes. Normally, a colour in an image is represented utilizing 3 bytes. With byte compaction, the proposed attack in this survey was able to stand for the colourss utilizing 2 bytes. Less figure of bytes used means less file size which is necessary in image compaction. The consequences they were able to garner shown the effectivity and efficiency of their proposed attack. However, since they merely used averaging in the block optimisation stage, the resulting image is non assured to be the optimum imaged that can be produced.
Harmony Search Algorithm of Geem et. Al is a new metaheuristic officially defined in 2001 [ 5 ] [ 6 ] . Even though it is a comparatively new optimisation algorithm, the effectivity and efficiency on optimisation jobs has already been shown in old surveies. With regard to the benchmark jobs like Traveling Salesman Problem, Harmony Search has outperformed the other antecedently bing optimisation algorithm which includes Genetic Algorithm, Ant Colony Optimization and Particle Swarm Optimization. The public presentation of Harmony Search has been tested non merely on benchmark jobs. This new metaheuristic has besides outperformed other optimisation algorithm when tested on existent universe job. Water distribution web is one of those jobs. Harmony Search was able to happen a more optimum solution than Genetic Algorithm in footings of cost. Having shown the potency of Harmony Search Algorithm in work outing optimisation job, it would be interesting to utilize this metaheuristic as an implicit in algorithm for other Fieldss particularly image compaction.
Geem and Williams [ 7 ] has applied Harmony Search on ecological optimisation specifically Maximal Covering Species Problem ( MCSP ) . This is an ecological preservation job that tries to continue species and their home ground. MCSP attempts to happen the maximal figure of species while restricting the figure of package. The algorithm was tested on what they call Oregon information which consists of 426 species and 441 packages. To be able to accommodate the algorithm to the job, they modified the construction of HS. Since their determination variable has merely two possible values, they omitted the pitch accommodation operation. Based on the consequences they were able to garner, they have shown the feasibleness of HS on ecological optimisation. Not merely was it executable, HS has outperformed Simulated Annealing, which is another metaheuristic, in work outing this job.
3 Statement of the Problem
Image compaction has been used in doing digital transmittal faster and it enabled more informations to be stored on same sum disc. Through remotion of excess informations and unobtrusive alterations, it decreases the physical size of the image.
To be able to hold a successful image compaction, it must hold a colour decrease method and it can stand for colour utilizing fewer sums of bytes. Different optimisation algorithms have been used to be able to cut down the figure of different colourss in an image with minimum effects to the quality of the image.
Harmony Search Algorithm, a comparatively new optimisation algorithm, has been used in different jobs that require optimisation. In this survey, the research worker aims to research the feasibleness of Harmony Search Algorithm as the underlying algorithm in accomplishing image compaction.
Nov 13 – 16, 2010
Research and reading of diaries, articles and published surveies about the assigned algorithm ( HSA ) and subject of involvement.
Nov 17 – 19, 2010
Making of study about the 3 proposed subjects
Nov 22 – 25, 2010
More research and reading about the 3 proposed subject and devising of presentation
Dec 1 – 3, 2010
Revision of study on the subject proposal for the chosen subject.
Dec 4 – 14, 2010
More research and reading on the chosen subject and assigned algorithm,
Making of the first portion of the proposal ( Introduction, Literature Review, etc. )
Dec 15, 2010 – Jan 4, 2011
Execution of proposed attack with the assigned algorithm