A Hybrid Image Compression Technique Using Wavelet Transformation – MFOCPN and Interpolation.
In this paper an interpolation method is proposed for compression technique. The method used is the localizing of spatial and frequency correlation from wavelets. Modified Forward Only Counter Propagation Neural Network (MFOCPN) is used for the classification and functional task. The wavelet based technique decomposes the lower sub band consisting of non significant coefficients and are eliminated. The significant smooth and sharp coefficients are found using interpolation methods. Here a new technique is proposed called the cosine interpolation, which is an alternative to the nearest neighborhood interpolation method. This methodology of interpolation proved to be an efficient approach for mapping all significant coefficients and thus resulting in improved quality. Hence the comparison is made between nearest neighborhood interpolation and cosine interpolation. The experimental results are tested on various standard images, where these results yield a better PSNR value compared with the existing nearest neighbor interpolation method.
