A New Modified Fast Fractal Image Compression Algorithm in DCT Domain

A New Modified Fast Fractal Image Compression Algorithm in DCT Domain

Jiji J.S.

PG Scholar School of CSE

Mar Ephraem College of Engineering and Technology Elavuvilai, Marthandam, India

Herlin L.T Assistant Professor School of CSE

Mar Ephraem College of Engineering and Technology Elavuvilai, Marthandam, India

Ashwin G Singerji Assistant Professor School of CSE

Mar Ephraem College of Engineering and Technology Elavuvilai, Marthandam, India

Sonal Wilson Pillai PG Scholar

School of CSE

Mar Ephraem College of Engineering and Technology Elavuvilai, Marthandam, India

Abstract- Medical Image Processing is the technique of analysing medical image such as CT, MRI for diagnostic and treatment purpose. The role of image compression is vital in domain of tele medicine for storage and transmission of medical data. In this project transform domain based compression algorithm is employed for the efficient compression and transfer of medical data. The Discrete Cosine Transform (DCT) based fractal image compression algorithm is employed for real time CT medical image. The result is validated by performance metrics Peak Signal to Noise Ratio (PSNR) and compression ratio. The algorithm was developed in MATLAB 2010a.

Keywords Peak Signal to Noise Ratio (PSNR), Quadtree Algorithm, Fractal Image Compressor, Compression Ratio(CR), Discrete Cosine Transform(DCT), Inverse Discrete Cosine Transform (IDCT).

1. INTRODUCTION

   Fractal image compression is one of the technique in image compression that uses self-similarity in images. It has much interest due to its promise of high compression ratios at good decompression quality and it has the advantage of very fast decompression. Another advantage of fractal image compression is its multi resolution property and it is possible to zoom-in on sections of the image. Image compression addresses the problem of reducing the amount of data required to represent a digital image with no significant loss of information. This field significantly grows through the practical application of the theoretic work, when the probabilistic view of information and its representation transmission and compression.

   Amit Kumar Biswas et al proposed Fractal image compression algorithm to encode the image. This algorithm enhance the reconstructed quality of image with high compression ratio [1]. Sharmila et al proposed (DCT) Discrete Cosine Transform to compress the image. The compression ratio of color image is calculated using PSNR value. It also reduce the noise of the decompressed images [2]. Chandan Singh Rawat et al proposed the (PSNR) Peak Signal to Noise Ratio and (SSIM) Structural Similarity Index

   Algorithm to compress the image. The color image is compressed using DCT in to similar blocks [3]. Gowri Sankar Reddy et al proposed an image compression algorithm based on uniform thresholding, using (BBP) Bits Per Pixel and (PSNR) Peak Signal to Noise Ratio algorithm. The main advantage is storage capacity is saved. [4]. Vinaysahu et al proposed an approach using DTCWT with high level filtering technique using PCA (Principle Component Analysis) algorithm. The performance was measured by parameters like MSE and PSNR which gives better results [5]. Kozhemiakin et al proposed the compression ratio prediction technique used for noisy and almost noise-free remote sensing images using DCT Algorithm. The methodology is used to describe iterative for CR (Compression Ratio) which does not require any compression or decompression histogram values V that is too smaller than image size [6].

   Krishna et al proposed (DCT) Discrete Cosine Transform and (MSVQ) Multistage Vector Quantization to increase the image compression ratio. This methods also reduces the storage size of images [7]. Mayuri A. chavan proposed 2D integer wavelet transform of medical images to reduce the energy of Sub-bands. The global and local symmetries is based on 3D scalable lossless compression for medical image. The 2D integer wavelet transform is used to decorrelate the data [8]. Leila Makkaoui et al proposed image compression of (WCSNs) Wireless Camera Sensor Networks to decrease the energy sensors and to maintain a long Network lifetime. The Experimental of DCT based image compression algorithm allows the efficient trade-off between energy consumption and image distortion [9]. Tilo Ochotta et al proposed fractal image compression which is better when compared to other fractal methods. The total rate of partition is encoded with region edge map based on uniform partitions. Fractally the image will be reconstructed by sharper edges and contrast than JPEG 2000 from mild blocking artifacts [10].

2. MATERIALS AND METHODS

   A new method for fractal image compression is used to reduce the encoding time. The analysis is done on different parameters such as encoding time, compression ratio, SSIM and PSNR. The most important point is that the encoding time was decreased by using predefined values instead of searching the contrast scaling factor. Image compression address the problem of reducing the amount of data required to represent a digital image with no significant loss of information.

   The steps used in the image compression algorithm is as follows

   • Read the input image to be compressed.

   • Partition the image into range block of size X × X and select the corresponding domain blocks of size 2X × 2X, which contained range block located at the center of the domain blocks.

   • Apply Discrete cosine transform on Range and Domain Blocks.

   • DCT is split into domain pool and range pool.

   • Apply Quad tree Algorithm for encoding.

   • Calculate Compression Ratio.

   Figure 1: Proposed Fractal Image Compressor

   At the first step of fractal coding, the image is partitioned into two non overlapping range blocks of size X × X where, X is a predefined parameter. Then a set of domain blocks is created from original image, taking all square blocks of size 2X × 2X with integer step L, in horizontal and vertical directions. Three new domain blocks are Created by rotating clockwise in 90º, 180º and 270º. These three domain

   blocks and the original domain block are mirrored. In addition to the original domain block, we have 7 new domain blocks. These 7 new domain blocks are added to the domain pool. After constructing the domain pool, each range block will be partitioned. we must select the best domain block from domain pool and find an affine transformation that maps the selected domain block to it with minimum distance. The mentioned distance between a range block, R, and a decimated domain block, D, both with n pixels is defined as follows:

   (1)

   Where the best coefficient s and o are ,

   (2)

   (3)

   Where <, >, R, D, are inner product, range block, domain block, mean of R and mean of D respectively. Because of high computational cost of (2), it is convenient to search S across a pre-sampled set of [0,1], instead of calculating (2). The process of encoding described above can be done entirely in one step by ignoring the error values of mappings.

   Figure 2: Position of range block and corresponding domain block.

   Discrete Cosine transform is applied in the range and domain block to separate the domain and range pool. Let N be the positive integer. The one dimensional DCT of order n is shown by n x n matrix M. It is given as,

   (4)

   The first parameter is the compression ratio, expressing how much less there are the coded data than the data of the original raw image.

   (5)

   PSNR is a metric for theratio between the maximum possible power of a signal and the power of the decoded image with reduced fidelity of its representation. It is most commonly used to measure the quality of decoded signals or images. It is defined most easily via the mean square error (MSE) that is defined for two images B and B decoded as follows,

   The first the correlation coefficient between x and y, which represents the degree of linear correlation between x and y, and the dynamic range is between 21 and 1. sx and sy can be considered as an estimate of the contrast in x and y. In practice, we use the metric described above, the image is windowed equally. Then, for each window, SSIM is used to find the average SSIM as follows,

   PSNR is defined as

   (6)

   (11)

   Where X and Y are the original and the de-noised images, respectively, M is the number of the local windows in the image, xj and yj are the image contents at the jth local window.

   (7)

   Here, MAX (B) is the maximum value of the numerical representation of a pixel luminance or chrominance component. Mostly, 8-bit values are used, and in this case, MAX (B) = 5255.

   The SSIM index can be used to measure the similarity of two images. This index combines three different metrics. Let x5{xi|i51, 2, 3,, N}, and y5v {yi|i51,2, 3,, N} be the original and the test image, respectively. SSIM is defined as,

   (8)

   Where,

   (9)

   The SSIM indicates any distortion as a combination of three different factors: loss of correlation, luminance distortion and contrast distortion. In other words, Q in equation (7) can be rewritten as a product of three components

   (10)

3. RESULTS AND DISCUSSIONS

   The first column represent the input human MRI image. Magnetic resonance imaging (MRI) is a medical imaging technique used in radiology to form pictures of the anatomy and the physiological processes of the body in both health and disease.

   The second column represent the compressed image of the given input image. Data compression is the process of encoding data using a representation that reduces the overall size of data. Digital image compression is a method for reducing the total number of bits required to represent an image.

   The third column represent the decompressed image. Decompression is accomplished by applying the inverse of each of the preceding step. Thus, the process starts with entropy decoding and proceeds to convert the run lengths to a sequence of zeros and coefficients. Coefficients are dequantized and Inverse Discrete Cosine Transform (IDCT) is performed to retrieve the decompressed image.

   Figure 3. Output results of Abdomen CT/MRI images

   Table I describes the performance matrix of Discrete Cosine transform with Elapsed time

   TABLE: PERFORMANCE MATRIX OF DISCRETE COSINE TRANSFORM

   Image Details	Time taken for compression	Time taken for decompressi on	PSNR	Compressio n Ratio
   1	3.9320	51.266324	22.54428	7.973066
   2	3.799529	46.242968	22.67624	8.786164
   3	3.237347	42.388705	22.37382	9.211948

4. CONCLUSION

A new method for fractal image compression is used. Different performance measures such as encoding time, compression ratio, SSIM and PSNR were analyzed. The most important point is that the encoding time was decreased by using predefined values instead of searching the contrast scaling factor. In addition, our analysis showed that domain pool reduction has a good performance for pool sizes less than 200, allowing shorter encoding times.

The algorithm has been verified with some well-known images, and the results have been compared with the state-of- the-art algorithms. The experiments show that the proposed algorithm has considerably lower encoding time, compared to the other algorithm with approximately the same quality of the encoded images. Also, the proposed method is better than the previous method based on entropy.

ACKNOWLEDGMENT

The authors we would like to acknowledge the support provided by DST under IDP scheme (No: IDP/MED/03/2015). We thank Dr Sebastian Varghese (Consultant Radiologist, Metro Scans & Laboratory, Trivandrum) for providing the medical CT/MR images and supporting us in the preparation of manuscript.

REFERENCES

1. Amit Kumar Biswas, Sanjeev Karmakar, Sanjay Sharma and Manoj Kumar Performance of Fractal Image Compression for Medical Images: A Comprehensive Literature Review International Journal of Applied Information Systems (IJAIS) ISSN: 2249-0868 Foundation of Computer Science FCS, New York, USA Volume 8 No.4, February 2015.

2. K. Sharmila, K. Kuppusamy An Efficient Image Compression Method using DCT, Fractal and Run Length Encoding Techniques International Journal of Engineering Trends and Technology (IJETT) Volume 13 Number 6 Jul 2014.

3. ChandanSinghRawat and Sukadev Meher A Hybrid Image Compression Scheme using DCT and Fractal Image Compression The International Arab Journal of Information Technology, Vol. 10, No. 6, November 2013.

4. D. Gowri Sankar Reddy1, V.C. Veera Reddy Image compression Based on Uniform Thresholding and Fixed Point Binary Scaling International Journal of Advanced Research in Computer and Communication Engineering Vol. 5, Issue 4, April 2016.

5. VinaySahu and Gagan Sharma A Hybrid Approach of Image Fusion Using Modified DTCWT with High Boost Filter Technique IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.3, March 2016.

6. R. A. Kozhemiakin1, A. N. Zemliachenko, V. V. Lukin1, S. K. Abramov, B. Vozel An approach to prediction and providing of compression ratio for DCT_based coder applied to remote sensing images 9 (2016) 2229.

7. V.Krishna, Dr.V.P.C.Rao, P.Naresh, P.Rajyalakshmi Incorporation of DCT and MSVQ to Enhance Image Compression Ratio of an image International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 03 Issue: 03 | Mar-2016

8. Mayuri A.Chavan Loss Free Compression of Medical Images International Journal Of Research in Engineering, Science and Technologies (IJRESTs) vol: 1 No: 6 Mar 2016

9. Leila Makkaoui, Vincent Lecuire and Jean-Marie Moureaux Fast Zonal DCT-based image compression for Wireless Camera Sensor Networks 2nd International Conference on Image Processing Theory, Tools and Applications, IPTA 2010, Jul 2010, Paris, France. IEEE, pp.126-129, 2010.

10. Tilo Ochotta and Dietmar Saupe EDGE-BASED PARTITION CODING FOR FRACTAL IMAGE COMPRESSION University of Konstanz Computer and Information Science Box M697 78457 Konstanz, Germany, c; 29(2004), 2C.-S.63-83.