This MSc thesis describes an image compression method, which is based on a technique that has not been applied widely in this area.
Using the singular value decomposition an optimal estimation can be made of a given image-block, which can be stored on much less space. The efficiency of the method highly depends on the size and content of this block. Therefore, the first step of the method is to split the frame into smaller blocks using an adaptive process. On the richly detailed or random-like parts of the picture the smaller blocks - while on the homogenous or characteristic structured areas the larger blocks are more effective.
The singular vectors obtained from the decomposition can be coded further. In the next step with a linear transformation the bit rate can be greatly reduced while little loss of quality.
The paper presents the mathematical tools used, describes the steps of the method and the MATLAB program (with the user interface) written for testing, and shows the results of the algorithm.