An efficient image compression technique by utilizing wavelet transform

Abstract

Image compression is very important for efficient transmission and storage of images. Various image compression techniques are employed to store or transmit image files to save storage space without any appreciable loss of information. But the PSNR rate and the compression ratio of these techniques are poor. The main goal of this thesis is to improve the PSNR rate and compression ratio. Accordingly, this thesis has made two significant contributions. The first contribution introduced a unique image compression technique by pooling Single Value Decomposition (SVD) and optimized Discrete Wavelet Transform (DWT). In the initial stage, the input image is subjected to decomposition by SVD. The decomposed image thus obtained is restored via inverse SVD. The restored image transformed by applying DWT and the optimization are performed by means of an optimization technique called genetic algorithm (GA). The outcome of the optimized DWT process will be given as input to the Huffman encoding and decoding. Finally, the original image is retained by means of Inverse Discrete Wavelet Transform (IDWT) process to get back the image. Simulation studies are performed on a MATLAB platform. The performance analysis is performed using Peak Signal Noise Ratio (PSNR), Structural Similarity Index Measurement (SSIM), Mean Square Error (MSE), and Compression Ratio (CR).The PSNR value for GADWT is 61.7864 where as that of DWT is 49.81945 when applying on 256*256 size images. The MSE 0.58683 is very small when comparing the other techniques. Hence the GA-DWT technique behaves much better than the other techniques. The second contribution offered the Dual-Tree Complex Wavelet Transform (DTCWT) Compression Technique to solve the problems in the compression system. DTCWT technique includes both DWT and CWT. At first, the input image undergoes SVD compression technique so that a few singular values can be eliminated by means of a predefined threshold. Subsequently, Inverse SVD is performed to restore the input image. This restored image is given as the input to the DTCWT. The result obtained from the DTCWT process will then be given for the Modified Huffman encoding and decoding process. Finally, to restore the image, the inverse transform is performed. The proposed technique performance is analyzed by using benchmarked images and the result has been compared with the conventional techniques in terms of performance metrics. The PSNR value for DTCWT is 76.7367 where as that of GA-DWT is 61.7864 when applying on 512*512 size images. The MSE 0.00158 is very small when comparing the other techniques. Hence the DTCWT technique performs much better than the other techniques. newline