The value of simplicity : image browsing techniques and tests

Sammanfattning: Image browsing is involved in many different types of tasks, such as interacting with a desktop and file system, viewing digital photo albums, etc. Different techniques are available for different types of tasks, but results from image browsing tests have been varied.The work of this thesis includes a literature survey, describing the most common image browser features and mechanisms, and summarizing research on the topic. The survey is followed by two image browsing studies, comparing an iconic interface to a zoom-and-pan and a fisheye interface. One study was made on a regular 19" display, and the other one on a 50" display with a 3D mouse. Both studies showed that none of the zoom-and-pan and the fisheye interfaces were more efficient than the simple iconic interface. The iconic interface was also the one most preferred by users. The problem with conflicting research results in this field is addressed in a summary of layout and presentation factors. There are a number of things, besides just interaction style, that affects usability. A lot remains to be investigated, and the analysis presented in this thesis gives an idea of some things to consider for future research. The two final studies of this thesis were parts of an interface design proposal for an oil analysis interface. One experiment showed that for reading single graphs, multiple diagrams with only one line graph in each are more efficient than one single chart with several line graphs. The final study of this thesis is a comparison of two graph tables, one distortion-oriented, and one overview+detail version. A theoretical analysis showed that for certain task types the static overview+detail interface would be more efficient than the distortion-oriented one, which was confirmed by the statistical analysis of data from the experiment. Users preferred the static overview+detail interface, mainly because they did not like that the view changed.The common experience of this work is that simplicity is highly valuable in many aspects. A simple interaction technique, such as an iconic interface, is often more efficient and more appreciated by users, than more advanced techniques. Reading single line graphs is more efficiently done in single small diagrams than in one large common diagram. Simple static modes are by many users preferred to dynamic modes. And finally, simple usability tests with a minimized number of confounding variables could explain and reduce the problem with conflicting results. Hence, the main conclusion and message of this thesis goes for both interface design and testing: Keep it simple.