ABSTRACT

The use of digital images has reached almost all the areas of our daily life. We often take pictures with our mobile phones or digital cameras, 3D movies are released in cinemas, computerized tomographies are required in hospitals, etc., being all of them in digital format. Moreover, data generated by these images must be processed, stored and, eventually, retrieved and displayed. Therefore, a matter of significant importance in digital image transmission is to design fast and reliable algorithms to deal with this increasing amount of information.

There are several techniques to handle data obtained from digital images but we will focus on progressive transmission patterns, which will arrange them so that they can be seen from the very beginning of the transmission process, where image quality will improve as more data are being received. An interesting aspect of progressive transmission is that a great level of effective compression can be achieved indirectly as the receiver can stop data transmission when enough detail level has been reached, or when images become irrelevant. Additionally, being this the main aim of our research, if the receiver is able to interact with the transmission process in order to select regions of interest (ROI) when he/she realizes that some region of the image is relevant, only data concerning that ROI will be sent.

An algorithm of lossy adaptive encoding for 2D and 3D digital images based on singular value decomposition (SVD) is proposed in our research and its efficiency will be studied, compared and discussed. Adaptive encoding is useful for progressive transmission of images and an algorithm for reconstruction will be developed, as well, from that encoding. Moreover, that encoding allows the progressive transmission of ROIs, either sequentially or several of them at the same time, at any step during the transmission process by making simple changes in the transmission and reconstruction algorithms, thus, avoiding re-encoding and redundancy in data transmission. 

Finally, we have considered digital video sequences as a 3D image where a combination of wavelets and SVD has been applied to each frame in order to develop an efficient simple codec.