Lossless compression of images with specific characteristics

The compression of some types of images is a challenge for some standard compression techniques. This thesis investigates the lossless compression of images with specific characteristics, namely simple images, color-indexed images and microarray images. We are interested in the development of complete compression methods and in the study of preprocessing algorithms that could be used together with standard compression methods. The histogram sparseness, a property of simple images, is addressed in this thesis. We developed a preprocessing technique, denoted histogram packing, that explores this property and can be used with standard compression methods for improving significantly their efficiency. Histogram packing and palette reordering algorithms can be used as a preprocessing step for improving the lossless compression of color-indexed images. This thesis presents several algorithms and a comprehensive study of the already existing methods. Specific compression methods, such as binary tree decomposition, are also addressed. The use of microarray expression data in state-of-the-art biology has been well established and due to the significant volume of data generated per experiment, efficient lossless compression methods are needed. In this thesis, we explore the use of standard image coding techniques and we present new algorithms to efficiently compress this type of images, based on finite-context modeling and arithmetic coding.