IMAGE COMPRESSION USING TRANSFORM BASED METHODS

Abstract

Image coding is a compression technique used for compressing the image by removing redundant information between different resolutions from the image. The advantages of image compression are we can save data size and hence transmission time. The image compression applications expect the algorithms with low complexity and processing time should be less. In this thesis work, a new proposed method for the image compression applications is implemented. The proposed method is QTBP (Quad tree based partitioning), which is based on block quantization. In this method image compression is achieved from lossy to lossless manner and for partitioning process of image, a sets of type S are used. This method is implemented because of its lower complexity (because here complex hierarchical tree structures are not used) and higher compression ratios. It uses a recursive set partitioning (quadrisected) procedure to sort subsets of wavelet coefficients by maximum magnitude with respect to thresholds that are integer powers of 2.The idea used in QTBP is it exploits two fundamental characteristics of the transformed image, energy clustering in frequency and space and a well defined hierarchical structure. In image compression applications choosing of transform based method is very important, from .past few years wavelet transforms are treating as successful method in various applications of image compression. Wavelet transforms are efficient transform methods because of its multiresolution nature by which we can decompose and analyse the image in different resolution or sub band levels. Generally conventional wavelet filters have floating point coefficients, which may leads to lossy compression, in order to avoid loss of data or information we choosen integer wavelet transform, which doesn't contain floating point coefficients. It is found that wavelet based methods give better results than discrete cosine transform based method for image compression applications. It is also found that proposed method QTBP is giving better results, by which higher compression ratios are achieved, so it is best suited for compressing images. The results of QTBP are compared to the state of the art methods like JPEG 2000 and SPIHT.