Image compression using modified set partitioning in hierarchical trees algorithms reducing memory space

Abstract

Images need resources for transmission and storage. So in order to minimize these requirements, image compression is very advantageous. For compression of data like images, wavelet transformation is found very effective and extremely powerful. Embedded Zerotree Wavelet (EZW) transform and Set Partitioning In Hierarchical Trees (SPIHT) are image compression techniques that show good quality of image after compression with high PSNR transmitting image in a progressive way. EZW algorithm is based on zerotrees while spatial orientation trees are partitioned in SPIHT algorithm in such a way that, insignificant coefficients are kept together in same subsets. The encoder decides for partitioning on the bases of binary decisions and transmits them to the decoder. The set partitioning process is followed by refinement process that refines the significant coefficients. This set partitioning in the form of subsets is found extremely effective and significant information is compressed immensely. In Discrete Wavelet Transform (DWT), a major issue is extension of image border. In ideal cases distortion should not be introduced when image border is extended under compression. This research presents the performance of several wavelet bases in Modifies Set Partitioning In Hierarchical Trees Algorithm (MSPIHT). Two types of wavelet bases are tested for MSPIHT algorithm i.e. orthogonal and biorthogonal wavelet bases. The orthogonal wavelet bases do not allow the simultaneous processing of two extremely significant properties needed for image compression i.e. orthogonality and symmetry. So biorthogonal wavelet bases are optimally suited for image compression using MSPIHT algorithm because they are capable of producing symmetric extension. These wavelet bases produce better results for low frequency images. The MSPIHT represents a more efficient implementation of the Set Partitioning In Hierarchical Trees algorithm by using variable thresholds to sort the List of Insignificant Pixels (LIP) and the List of Insignificant Sets (LIS) of Said’s SPIHT algorithm. The goal of this research is to compare the performance of MSPIHT algorithm with that of Embedded Zerotree Wavelet (EZW) algorithm and SPIHT algorithm and to use the coefficients of several wavelet filters instead of using just one wavelet filter for image compression using MSPIHT coding and to use variable thresholds for sorting the subsets and pixels of original SPIHT algorithm. The results of using these wavelet bases are compared on the basis of Compression Ratio (CR) and Peak Signal to Noise Ratio (PSNR). The research shows that use of biorthogonal wavelets bases is better than orthogonal wavelet bases. Out of biorthogonal wavelets, bior 4.4 shows good results in MSPIHT coding also the MSPIHT shows better results as compare to Said’s SPIHT and Shapiro’s EZW algorithm.