ENGLISH

Abstract

Multimedia digital libraries involve diflerent types of data objects such as text, audio, video and images. Most of these data objects are very large in size and accessing them in a distributed environrnent causes a transmission delay due to the vast amount of network traffic. Compressing these data objects before transmission, can reduce the response time, although it would mean a reduction in the quality of the output data. If the application doesn't demand a high quality output, data compression can be an acceptable means of reducing transmission time over the internet or other distributed environments. The loss of quality would be proportional to the amount of compression applied to individual data objects. Therefore different quality levels (browsing levek) can be achieved depending on how the data will be used. A lower quality level could be used for general browsing of data whereas a higher quality level could be used where the output data has to be further processed and analysed. The behaviour of images when coinpressed using various conlpression techniques was studied and it was observed that different images require different amounts of compression to reach the same quality level. This result allowed us to classify images into different classes, based on their compression behaviour. In this paper, we identify a set of rules to calculate a near optimal compression ratio to achieve a given level of image quality. We also explain how this set of rules can be incorporated i ~ t odig ital libraries with nlultiple levels of browsing, to achieve a faster response time. http://dx.doi.org/10.14429/dbit.17.6.3244