The main objective of this research is to design and implement novel spectral-spatial processing techniques for hyperspectral image analysis and applications. Although the high dimensionality of hyperspectral image data makes its transmission and storage difficult, the uncompressed data format is still preferred as it avoids compression loss which may degrade classification accuracy. In this thesis, a quality-assured lossy compression scheme based on a modified three dimensional discrete cosine transform is proposed. This novel technique is demonstrated to maintain the integrity of hyperspectral data without degrading the classification accuracy. Furthermore, this work has led to the creation of an effective spectral feature extraction technique which uses curvelet transform and singular spectrum analysis. In addition to this, an original classification framework which combines joint bilateral filtering and an improved sparse representation classifier is presented. Experimental results show that the proposed methodologies outperform most of the state-of-the-art feature extraction and classification techniques commonly employed in the hyperspectral community. This work also demonstrates that hyperspectral imaging combined with advanced signal processing is an effective technology for food quality control applications. For example, when applied to the challenge of performing hyperspectral imaging-based meat quality assessment, the techniques proposed in this work are shown to provide a more effective solution than conventional visible and near-infrared spectroscopic technology. Finally, this thesis provides the first set of results of assessing the quality of beef and lamb samples using an improved data regression technique. To sum up, the outcome of this thesis advances the hyperspectral imaging community by proposing several novel methodologies, and extensive experiments have been conducted to demonstrate their superiority.