There is a explosion in the number of devices that can create (digital cameras, scanners, DV cams etc.), share (Web, multimedia email etc.) and consume (inexpensive PCs, PDAs etc.) multimedia objects. The file size of the multimedia images place an enormous strain on the available resources. For example slow clients experience intolerable delays in accessing images while clients using expensive networks pay inordinately more resources to access images. Congested web servers use more of their scarce network bandwidth for multimedia images.
The goal of this research effort is to customize multimedia objects so that they are appropriate in more scenarios than they are in the original form. We utilize transcoding as the enabling technology. Transcoding is a transformation that is used to convert a multimedia object from one form to another (frequently trading off object fidelity for size). For transcoding to be useful, we need to understand the tradeoff characteristics: the information quality loss, the computational overhead required in computing the transcoding and the potential space benefits.
In order to answer the question whether transcoding operations will be worthwhile, we need to understand the nature of typical internet images and their transcoding characteristics. We focus our attention on transcodings that customize an image for file size savings.
We analyzed the characteristics of images available on the web [multimedia00]. We showed that most GIF images accessed on the Internet in a representative trace are small; about 80% of the GIF images are smaller than 6 KBs. On average, JPEG images are larger than GIF images; about 40% of the JPEG images are larger than 6 KBs.
We also established the characteristics of popular image transcoding operations. We showed that for JPEG images, the JPEG compression metric and a transcoding that reduces the spatial geometry are productive transcoding operations. We also showed that, unless proper transcoding parameters are chosen, traditional transcoding techniques such as thumbnailing of GIF images have the potential of actually increasing the output image size.
Based on the previous results, to illustrate for one specific case, we characterized the information quality tradeoffs, the computational requirements and the potential space gain of a transcoding that changes the JPEG compression metric.[usits99] We presented techniques to quantify the size tradeoff characteristics and the effects on image quality for transcoding JPEG images. We analyzed the characteristics of images available in typical Web sites and explored how we can perform informed transcoding using the JPEG compression metric. We presented the effects of this transcoding on the image storage size and image information quality. We also presented ways of predicting the computational cost as well as potential space benefits achieved by the transcoding.
Mobile applications increasingly depend upon multimedia information originating in the web and attempt to access this data over wireless networks that are more expensive and slower than typical wireline internet access from desktops. Transcoding is an important technique that can allow network proxy servers to customize the delivered object size for the network bandwidth available on the ``last hop'' to the mobile client. We exploit transcoding technology that we had previously developed [usits99] that characterized the quality versus size tradeoffs in transcoding JPEG images. This technology supplies more information for use in transcoding policy decision making. We evaluate the performance benefits of incrementally incorporating this information in a series of transcoding policies.
We demonstrated that it is possible to use informed transcoding techniques to balance the need for good quality of multimedia content while reducing consumed network bandwidth and server CPU overhead. We showed that policies that aggressively transcode the larger images can produce images with Quality factor values that closely follow the un-transcoded base case while still saving as much as 150 KB. A transcoding policy that has knowledge of the characteristics of the link to the client can avoid as many as 40\% of (unnecessary) transcodings.
The ability of a web service to provide low-latency access to its contents is constrained by available network bandwidth. It is important for the service to manage available bandwidth wisely. While providing differentiated quality of service (QoS) is typically enforced through network mechanisms, we introduce a robust mechanism for managing network resources at the application level [infocom00]. We use transcoding to allow web servers to customize the size of objects constituting a web page, and hence the bandwidth consumed by that page, by dynamically varying the size of multimedia objects on a per-client basis. We leverage earlier work on characterizing quality versus size tradeoffs in transcoding JPEG images [usits99] to dynamically determine the quality and size of the object to transmit. We evaluated the performance benefits of incorporating this information in a series of bandwidth management policies. We developed metrics to measure the performance of our system. We used realistic workloads and access scenarios to drive our system.
The principal contribution of this work was the demonstration that it is possible to use informed transcoding techniques to provide differentiated service and to dynamically allocate available bandwidth among different client classes, while delivering a high degree of information content (quality factor) for all clients.