Hybrid JPEG Compression Using Histogram Based Segmentation by ijcsis


									                                                     (IJCSIS) International Journal of Computer Science and Information Security,
                                                     Vol. 8, No. 8, November 2010

Hybrid JPEG Compression Using Histogram Based
                     M.Mohamed Sathik1 , K.Senthamarai Kannan2 and Y.Jacob Vetha Raj3
                        Department of Computer Science, Sadakathullah Appa College, Tirunelveli, India
                         Department of Statistics, Manonmanium Sundaranar University, Tirunelveli, India

  Abstract-- Image compression is an inevitable solution for           lossy compression algorithms obtain high compression
image transmission since the channel bandwidth is limited and          ratios by exploiting human visual properties.
the demand is for faster transmission. Storage limitation is also              Vector quantization [4],[5] , wavelet transformation
forcing to go for image compression as the color resolution and        [1] , [5]-[10] techniques are widely used in addition to
spatial resolutions are increasing according to quality                various other methods[11]-[17] in image compression. The
requirements. JPEG compression is a widely used compression
                                                                       problem in lossless compression is that, the compression
technique. JPEG compression method uses linear quantization
and threshold values to maintain certain quality in an entire          ratio is very less; where as in the lossy compression the
image. The proposed method estimates the vitality of the block         compression ratio is very high but may loose vital
of the image and adapts variable quantization and threshold            information of the image. Some of the works carried out in
values. This ensures that the vital area of the image is highly        hybrid image compression [18]-[19] incorporated different
preserved than the other areas of the image. This hybrid               compression schemes like PVQ and DCTVQ in a single
approach increases the compression ratio and produces a                image compression. But the proposed method uses lossy
desired high quality output image.                                     compression method with different quality levels based on
  .                                                                    the context to compress a single image by avoiding the
                                                                       difficulties of using side information for image
  Key words-- Image Compression, Edge-Detection,                       decompression in [20].
 Segmentation. Image Transformation, JPEG,                                     The proposed method performs a hybrid
 Quantization.                                                         compression, which makes a balance on compression ratio
                                                                       and image quality by compressing the vital parts of the
                     I. INTRODUCTION                                   image with high quality. In this approach the main subject
             Every day, an enormous amount of                          in the image is very important than the background image.
 information is stored, processed, and transmitted digitally.          Considering the importance of image components, and the
 Companies provide business associates, investors, and                 effect of smoothness in image compression, this method
 potential customers with financial data, annual reports,              segments the image as main subject and background, then
 inventory, and product information over the Internet. And             the background of the image is subjected to low quality
 much of the online information is graphical or pictorial in           lossy compression and the main subject is compressed with
 nature; the storage and communications requirements are               high quality lossy compression. There are enormous
 immense. Methods of compressing the data prior to storage             amount of work on image compression is carried out both
 and/or transmission are of significant practical and                  in lossless [1] [14] [17] and lossy [4] [15] compression.
 commercial interest.                                                  Very few works are carried out for Hybrid Image
                                                                       compression [18]-[20].
         Compression techniques fall under two broad                           In the proposed work, for image compression, the
 categories such as lossless[1] and lossy[2][3]. The former            edge detection, segmentation, smoothing and dilation
 is particularly useful in image archiving and allows the              techniques are used. For edge detection, segmentation
 image to be compressed and decompressed without losing                [21],[22] smoothing and dilation, there are lots of work has
 any information. And the later, provides higher levels of             been carried out [2],[3]. A novel and a time efficient
 data reduction but result in a less than perfect reproduction         method to detect edges and segmentation used in the
 of the original image. Lossy compression is useful in                 proposed work are described in section II, section III gives
 applications      such      as      broadcast      television,        a detailed description of the proposed method, the results
 videoconferencing, and facsimile transmission, in which               and discussion are given in section IV and the concluding
 certain amount of error is an acceptable trade-off for                remarks are given in section V.
 increased compression performance. The foremost aim of
 image compression is to reduce the number of bits needed
                                                                                        II. BACKGROUND
 to represent an image. In lossless image compression
 algorithms, the reconstructed image is identical to the              A. JPEG Compression
 original image. Lossless algorithms, however, are limited
                                                                      Components of Image Compression System (JPEG).
 by the low compression ratios they can achieve. Lossy
                                                                      Image compression system consists of three closely
 compression algorithms, on the other hand, are capable of
                                                                      connected components namely
 achieving high compression ratios. Though the
 reconstructed image is not identical to the original image,                       Source encoder (DCT based)

                                                                300                             http://sites.google.com/site/ijcsis/
                                                                                                ISSN 1947-5500
                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                  Vol. 8, No. 8, November 2010

               Quantizer                                                           The DCT based encoder can be thought of as essentially
               Entropy encoder                                                     compression of a stream of 8 X 8 blocks of image samples.
Figure 2 shows the architecture of the JPEG encoder.                                Each 8 X 8 block makes its way through each processing
Principles behind JPEG Compression.                A common                         step, and yields output in compressed form into the data
characteristic of most images is that the neighboring pixels                        stream. Because adjacent image pixels are highly correlated,
are correlated and therefore contain redundant information.                         the ‘forward’ DCT (FDCT) processing step lays the
The foremost task then is to find less correlated                                   foundation for achieving data compression by concentrating
representation of the image. Two fundamental components                             most of the signal in the lower spatial frequencies. For a
of compression are redundancy and irrelevancy reduction.                            typical 8 X 8 sample block from a typical source image,
Redundancy reduction aims at removing duplication from                              most of the spatial frequencies have zero or near-zero
the signal source. Irrelevancy reduction omits parts of the                         amplitude and need not be encoded. In principle, the DCT
signal that will not be noticed by the signal receiver, namely                      introduces no loss to the source image samples; it merely
the Human Visual System (HVS). The JPEG compression                                 transforms them to a domain in which they can be more
standard (DCT based) employs the use of the discrete cosine                         efficiently encoded.
transform, which is applied to each 8 x 8 block of the                              After output from the FDCT, each of the 64 DCT
partitioned image. Compression is then achieved by                                  coefficients is uniformly quantized in conjunction with a
performing quantization of each of those 8 x 8 coefficient                          carefully designed 64 – element Quantization Table. At the
blocks.                                                                             decoder, the quantized values are multiplied by the
Image Transform Coding For JPEG Compression                                         corresponding QT elements to recover the original
Algorithm.                                                                          unquantized values. After quantization, all of the quantized
In the image compression algorithm, the input image is                              coefficients are ordered into the “zig-zag” sequence as
divided into 8-by-8 or 16-by-16 non-overlapping blocks,                             shown in figure 1. This ordering helps to facilitate entropy
and the two-dimensional DCT is computed for each block.                             encoding by placing low-frequency non-zero coefficients
The DCT coefficients are then quantized, coded, and                                 before high-frequency coefficients. The DC coefficient,
transmitted. The JPEG receiver (or JPEG file reader)                                which contains a significant fraction of the total image
decodes the quantized DCT coefficients, computes the                                energy, is differentially encoded.
inverse two-dimensional DCT of each block, and then puts
the blocks back together into a single image. For typical                                                AC01
                                                                                          DCC                                                   AC07
images, many of the DCT coefficients have values close to
zero; these coefficients can be discarded without seriously
affecting the quality of the reconstructed image.       A two
dimensional DCT of an M by N matrix A is defined as
         M -1 N-1
                             (2m +1)p      (2n +1)q
Bpq  αpαq   Amn cos                 cos            , 0  p  M-1
         m=0 n=0                2M             2N
                                                                0  q  N-1

                                                                                                                                               AC77
                                    αp =  1/2/M ,, 1=p0 M-1
                                               M p
                                                                                                         Figure 1 Zig-Zag Sequence
                                    αq =  1/2/N ,, 1=q0 N-1
                                              N q
                                                                                            The JPEG decoder architecture is shown if figure 3
                                                                                    which is the reverse procedure described for compression.
The DCT is an invertible transformation and its inverse is
given by
                                                                                    B. Segmentation
         M-1 N-1
                             (2m +1)p      (2n +1)q                                         Let D be a matrix of order m x n, represents the
Amn  αpαq        Bpq cos            cos            , 0  p  M-1
         m=0 n =0               2M             2N                                   image of width m and height n. The domain for Di,j is
                                                                0  q  N-1
                                                                                    [0..255] , for any i=1..m, and any j=1.. n.
                                                                                              The architecture of segmentation using histogram
 Where                                                                              is shown in figure 4. To make the process faster the high
                                                                                    resolution input image is down sampled 2 times. When the
                      αp =  1/2/M ,, 1=p0 M-1
                                 M p
                                                                                    image is down sampled each time the dimension is reduced
                                                                                   by half of the original dimension. So the final down sampled
                                                                                                                             m      n
                                                                                   image (D) is of the dimension 4 x 4 . The down sampled
                       αq =  1/2/N ,, 1=q0 N-1
                                  N q
                                                                                   image is smoothed to get smoothed gray scale image using
                                                                                    equation (1).
                                                                                    Si,,j = 1/9(Di-1,,j-1 +Di-1,,j +Di-1,,j+1 +Di,,j-1 +Di,,j +Di,,j+1 +Di+1,,j-1
                                                                                    +Di+1,,j +Di+1,,j+1 )                                               …(1)

                                                                              301                                   http://sites.google.com/site/ijcsis/
                                                                                                                    ISSN 1947-5500
                                                 (IJCSIS) International Journal of Computer Science and Information Security,
                                                 Vol. 8, No. 8, November 2010

          8 X 8 blocks

                                    FDCT             Quantizer                      Entropy                   Compressed
                                                                                    Encoder                   image data

                                                Quantizer                     Huffman
              Source image                       Table                         Table

                                          Figure 2. JPEG Encoder Block Diagram

Compressed               Entropy           Dequantizer                   IDCT
Image data               Decoder

                                            Quantizer                  Huffman                   Reconstructed
                                             Table                      Table                       Image

                                          Figure 3. JPEG Decoder Block Diagram

                                                                     Bi,j = (Si,j > L) and (Si,j < U)            …(5)

     Down             Smooth the             Histogram
    Sample              Image               Computation
  2 Times (D)            (S)                    (H)

      Up                Binary                Range
    Sample           Segmentation           Calculation
    2 Times                (B)               (L & U)

                   Figure -4 Classifier
 The histogram(H) is computed for the gray scale image(S).
The most frequently present gray scale value (Mh) is
determined from the histogram by equation (2) and is shown
as indicated by a line in figure 5.
                    Mh = arg{max(H(x))}                                                           Figure – 5

The background value of the images is having the highest
frequency in the case of homogenous background. In order             After processing the pixel values for background area is 1 in
to surmise background textures a range of gray level                 the binary image B. To avoid the problem of over
values are considered for segmentation. The range is                 segmentation the binary image is subjected to sequence of
computed using the equations (3) and (4).                            morphological operations. The binary image is eroded with
                  L = max( Mh – 30,0)                   …(3)         smaller circular structural element (SE) to remove smaller
                                                                     segments as given in equation (6).
                  U = min( Mh + 30,255)                 …(4)
                                                                                               B  BSE                                  …(6)
The gray scale image S is segmented to detect the
background area of the image using the function given in
equation (5)

                                                               302                                http://sites.google.com/site/ijcsis/
                                                                                                  ISSN 1947-5500
                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                    Vol. 8, No. 8, November 2010

Then the resultant image is subjected to morphological                      10. Quantize the DCT coefficients
closing operation with larger circular structural element as
                                                                            11. Discard lower quantized values based on the
given in equation (7).
                                                                                threshold value selected by the selector.
                        B  B  SE                       …(7)               12. Compress remaining DCT coefficients by Entropy
  III. PROPOSED HYBRID JPEG COMPRESSION.                                        Encoder

          The input image is initially segmented into                   The architecture of the proposed method is shown in figure
background and foreground parts as described in section                 6. The Quantization Table is a fixed classical table derived
II.B Then the image is divided into 8x8 blocks and DCT                  from empirical results. The Quantizer quantizes the DCT
values are computed for each block. The quantization is                 coefficients computed by FDCT. The classifier identifies the
performed according to the predefined quantization table.               class of each pixel by segmenting the given input image.
The quantized values are then reordered based on zig-zag                The selector and limiter works together to find the discard
ordering method described in section II A. The lower values             threshold limit. The entropy encoder creates compressed
of AC coefficients are discarded from the zig-zag ordered               code using the Code Table. The compressed image may be
list by comparing the threshold value selected by the                   stored or transmitted faster than the existing method.
selector as per the block’s presences identified by the                            IV. RESULTS AND DISCUSSION
classifier. If the block is present in foreground area then the
threshold is set to a higher value by the selector, otherwise a         The Hybrid JPEG Compression method is implemented

                                                            FDCT                         Quantizer

                                  Classifier               Selector                       Limiter

                                                          Code Table                     Entropy


                                     Figure 6. Hybrid JPEG compression method
lower value for threshold is set by the selector. After according to the description in section III and tested with a
discarding insignificant coefficients the remaining data are set of test images shown in figure 8. The results obtained
compressed by the standard entropy encoder based on the from the implementation of the proposed algorithms are
code table.                                                   shown in figures 7, 9,10 and table I. Figure 7.a shows the
                                                              original input image. In Figure 7.b the segmented object
Algorithm                                                     and background area is discriminated by black and white.
    1. Input High Resolution Color image.                     The compressed bit rates of the twelve test images are
                                                              computed and tabulated in table 1. The low quality (LQ) and
    2. Down sample the input image 2 times.                   high quality (HQ) JPEG compression is performed and the
    3. Convert the down sampled image to gray scale corresponding compression ratios(CR) and PSNR values
         image (G).                                           are tabulated. The PSNR is higher for HQ and CR is higher
                                                              for LQ. The Hybrid JPEG compression performs HQ
    4. Find histogram (H) of the gray scale image.            compression on main subject area and LQ compression on
    5. Find the lower (L) and upper (U) gray scale value background area thus the PSNR value at main subject area is
         of background area.                                  the same for Hybrid JPEG and HQ JPEG. Figure 9 shows
                                                              the comparison of normalized CRs of Hybrid JPEG and HQ
    6. Find Binary segmented image (B) from the gray JPEG, it is observed that almost all of the images are
         scale image (G)                                      compressed better than classical JPEG compression. Figure
    7. Up sample Binary image (B) two times.                  10 shows how well the compression ratio is increased than
                                                              the classical JPEG compression method.
    8. Divide the input image into 8x8 blocks
    9.   Find DCT coefficients for each blocks

                                                                  303                           http://sites.google.com/site/ijcsis/
                                                                                                ISSN 1947-5500
                                      (IJCSIS) International Journal of Computer Science and Information Security,
                                      Vol. 8, No. 8, November 2010

                                             a)Input Image

                               b)Segmented Main Subject Area
                                      Figure – 7 Input /Output

         Table -1 Compression Ratio and PSNR values obtained by Hybrid JPEG and JPEG

Image           LQ                  Hybrid             Hybrid/HQ                HQ
        CR1      PSNR1       CR2        PSNR2           PSNR@            CR3       PSNR3         CR2-CR3
 3      26.00        21.52   7.46       27.82            27.84           7.46        27.82         0.0000
 5      25.29        22.09   6.80       28.87            28.93           6.80        28.87         0.0004
 10     23.83        20.78   5.41       28.09            28.13           5.41        28.09         0.0031
 4      24.33        22.08   6.61       31.13            31.20           6.60        31.13         0.0068
 11     27.75        25.33   8.62       35.97            36.25           8.61        36.01         0.0198
 9      24.92        21.62   6.58       30.45            30.56           6.56        30.51         0.0204
 1      27.54        23.03   8.40       29.37            29.40           8.37        29.37         0.0276
 7      24.85        17.71   4.01       23.38            23.43           3.92        23.38         0.0911
 8      24.63        21.97   6.73       31.05            31.13           6.64        31.10         0.0921
 12     29.18        22.73   5.92       28.70            29.07           5.61        28.93         0.3111
 6      26.47        20.12   5.54       25.31            25.33           5.19        25.30         0.3449
 2      22.84        20.40   7.91       27.93            28.21           7.25        28.06         0.6541

                                                 304                             http://sites.google.com/site/ijcsis/
                                                                                 ISSN 1947-5500
                                                                                (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                Vol. 8, No. 8, November 2010

                                                       Figure – 8 Test Images (1-12 from Left to Right and Top to Bottom)


                   Normalized CR

                                                                                                                                            HQ Lossy


                                                           1   2   3    4       5   6     7         8          9   10 11 12

                                                       Figure – 9 Normalized Compression Ratio Obtained for Test Images
                                                                   Increased Compression Ratio by Hybrid JPEG Compression



                         Compression Ratio





                                                       1       2    3       4       5         6            7       8   9      10       11      12

                                                       Figure – 10 Increased Compression Ratios by Hybrid Compression.

                   V. CONCLUSION                                                                   the background area is lower in Hybrid JPEG method which
                                                                                                   is acceptable, since the background area is not vital. The
         The compression ratio of Hybrid JPEG method is
                                                                                                   Hybrid JPEG method is suitable for imagery with larger
higher than JPEG method in more than 90% of test cases. In
                                                                                                   trivial background and certain level of loss is permissible.
the worst case both Hybrid JPEG and JPEG method gives
the same compression ratio. The PSNR value at the main
subject area is same for both methods. The PSNR value at

                                                                                           305                                     http://sites.google.com/site/ijcsis/
                                                                                                                                   ISSN 1947-5500
                                                           (IJCSIS) International Journal of Computer Science and Information Security,
                                                           Vol. 8, No. 8, November 2010

                     ACKNOWLEDGEMENTS                                                   Exploring the Space of Realistic Distortions”, IEEE Transactions
                                                                                        on Image Processing, vol. 17, no. 8, pp. 1261–1273, Aug 2008.
         The authors express their gratitude to University                       [20]   Hong, S. W. Bao, P., “Hybrid image compression model based
Grant Commission and Manonmanium Sundaranar                                             on subband coding and edge-preserving regularization”, Vision,
                                                                                        Image and Signal Processing, IEE Proceedings, Volume: 147,
University for financial assistance under the Faculty                                   Issue: 1, 16-22, Feb 2000
Development Program.                                                             [21]   Zhe-Ming Lu, Hui Pei ,”Hybrid Image Compression Scheme
                                                                                        Based on PVQ and DCTVQ “,IEICE - Transactions on
                                                                                        Information and Systems archive, Vol E88-D , Issue 10 ,October
REFERENCES                                                                              2006
                                                                                 [22]   Y.Jacob Vetha Raj, M.Mohamed Sathik and K.Senthamarai
  [1]    Xiwen OwenZhao, Zhihai HenryHe, “Lossless Image                                Kanna,, “Hybrid Image Compression by Blurring Background
         Compression Using Super-Spatial Structure Prediction”, IEEE                    and Non-Edges. The International Journal on Multimedia and its
         Signal Processing Letters, vol. 17, no. 4, April 2010 383                      applications.Vol 2, No.1, February 2010, pp 32-41
  [2]    Aaron T. Deever and Sheila S. Hemami, “Lossless Image                   [23]   Willian K. Pratt ,”Digital Image Processing” ,John Wiley & Sons,
         Compression With Projection-Based and Adaptive Reversible                      Inc, ISBN 9-814-12620-9.
         Integer Wavelet Transforms”, IEEE Transactions on Image                 [24]   Jundi Ding, Runing Ma, and Songcan Chen,”A Scale-Based
         Processing, vol. 12, NO. 5, May 2003.                                          Connected Coherence Tree Algorithm for Image Segmentation”,
  [3]    Nikolaos V. Boulgouris, Dimitrios Tzovaras, and Michael                        IEEE Transactions on Image Processing, vol. 17, NO. 2, Feb
         Gerassimos Strintzis, “Lossless Image Compression Based on                     2008.
         OptimalPrediction, Adaptive Lifting, and Conditional Arithmetic         [25]    Kyungsuk (Peter) Pyun, , Johan Lim, Chee Sun Won, and Robert
         Coding”, IEEE Transactions on Image Processing, vol. 10, NO.                   M. Gray, “Image Segmentation Using Hidden Markov Gauss
         1, Jan 2001                                                                    Mixture Models”, ” ,IEEE Transactions on Image Processing,
  [4]    Xin Li, , and Michael T. Orchard “ Edge-Directed Prediction for                VOL. 16, NO. 7, JULY 2007
         Lossless Compression of Natural Images”, IEEE Transactions on
         Image Processing, vol. 10, NO. 6, Jun 2001.
  [5]    Jaemoon Kim, Jungsoo Kim and Chong-Min Kyung , “A
         Lossless Embedded Compression Algorithm for High Definition
         Video Coding” 978-1-4244-4291 / 09 2009 IEEE. ICME 2009
  [6]    Rene J. van der Vleuten, Richard P.Kleihorstt ,Christian
         Hentschel,t    “Low-Complexity          Scalable    DCT   Image
         Compression”, 2000 IEEE.
  [7]    K.Somasundaram, and S.Domnic, “Modified Vector Quantization
         Method for mage Compression”, ’Transactions On Engineering,
         Computing And Technology Vol 13 May 2006
  [8]    Mohamed A. El-Sharkawy, Chstian A. White and Harry
         ,”Subband Image Compression Using Wavelet Transform And
         Vector Quantization”, 1997 IEEE.
  [9]    Amir Averbuch, Danny Lazar, and Moshe Israeli ,”Image
         Compression Using Wavelet Transform and Multiresolution
         Decomposition”, IEEE Transactions on Image Processing, vol. 5,
         NO. 1, Jan 1996.
  [10]   Jianyu Lin, Mark J. T. Smith,” New Perspectives and
         Improvements on the Symmetric Extension Filter Bank for
         Subband /Wavelet Image Compression”, IEEE Transactions on
         Image Processing, vol. 17, NO. 2, Feb 2008.
  [11]   Yu Liu, Student Member, and King Ngi Ngan, “Weighted
         Adaptive Lifting-Based Wavelet Transform for Image Coding
         “,IEEE Transactions on Image Processing, vol. 17, NO. 4, Apr
  [12]   Michael B. Martin and Amy E. Bell, “New Image Compression
         Techniques Using Multiwavelets and Multiwavelet Packets”
         ,IEEE Transactions on Image Processing, vol. 10, NO. 4, Apr
  [13]   Roger L. Claypoole, Jr , Geoffrey M. Davis, Wim Sweldens
         ,“Nonlinear Wavelet Transforms for Image Coding via Lifting” ,
         IEEE Transactions on Image Processing, vol. 12, NO. 12, Dec
  [14]   David Salomon, “Data Compression , Complete Reference”,
         Springer-Verlag New York, Inc, ISBN 0-387-40697-2.
  [15]   Eddie Batista de Lima Filho, Eduardo A. B. da Silva Murilo
         Bresciani de Carvalho, and Frederico Silva Pinagé “Universal
         Image Compression Using Multiscale Recurrent Patterns With
         Adaptive Probability Model” , IEEE Transactions on Image
         Processing, vol. 17, NO. 4, Apr 2008.
  [16]   Ingo Bauermann, and Eckehard Steinbach, “RDTC Optimized
         Compression of Image-Based Scene Representations (Part I):
         Modeling and Theoretical Analysis” , IEEE Transactions on
         Image Processing, vol. 17, NO. 5, May 2008.
  [17]   Roman Kazinnik, Shai Dekel, and Nira Dyn , ”Low Bit-Rate
         Image Coding Using Adaptive Geometric Piecewise Polynomial
         Approximation”, IEEE Transactions on Image Processing, vol.
         16, NO. 9, Sep 2007.
  [18]   Marta Mrak, Sonja Grgic, and Mislav Grgic ,”Picture Quality
         Measures in Image Compression Systems”, EUROCON 2003
         Ljubljana, Slovenia, 0-7803-7763-W03 2003 IEEE.
  [19]   Alan C. Brooks, Xiaonan Zhao, , Thrasyvoulos N. Pappas.,
         “Structural Similarity Quality Metrics in a Coding Context:

                                                                           306                               http://sites.google.com/site/ijcsis/
                                                                                                             ISSN 1947-5500

To top