We conducted user studies to compar

We conducted user studies to compare our system with two Web image search engines, Google image search and Bing image search. In the following, we use engine A to represent one of the two engines, and engine B to represent the other engine. We justify our system by comparing the performance with Engines A and B, investigating the requirement of search-by-color-map, and inspecting the usability of the user interface.
6.1 Design Participants We recruit 30 volunteers, students from a university campus and our research lab, to take part in the user study. Their grades vary from freshman to graduate grade 3. Their ages range from 19 to 24. All participants are Web image search engine users. Data set We collect a large scale of images that come from Microsoft Bing image search for our system. For each image in the database, besides the text-based indexing, we compute two kinds of features offline, dominant colors and color map features, and we organize the images based on the dominant color using the text based technique for efficient dominant color filtering. Search tasks We compare the performance of our approach with the other two image search engines over a set of tasks. We collect the tasks that are conducted by the volunteers when they tried our system. We select a part of tasks (40 tasks) from those collected tasks. These tasks are given in descriptions that express the search intent, e.g., “the beach on the bottom left and the sky on the top right”. Sample tasks are listed as follows:
(1) White dog on grass; (2) Island in the sea; (3) Eiffel tower on the lawn;