Review of Human-Computer Interaction Issues in Image Retrieval [chapter]

Mohammed Lamine
2008 Advances in Human Computer Interaction  
www.intechopen.com Advances in Human-Computer Interaction 216 Review of Human-Computer Interaction Issues in Image Retrieval 217 images within a reasonable time. These tools are generally called image retrieval engines, or image retrieval systems. Different scenarios are possible for image retrieval. The most common scenario is the following: 1. The engine allows the user to create his query. It may be a text box in which the user enters keywords describing what he is searching. It may also
more » ... lve a set of images from which the user can choose several as examples. Other ways of creating the query are also possible, as we will see later in the chapter. 2. The user creates his query. 3. The engine searches by comparing the query against the images in the collection. 4. The engine displays the resulting images for the user. 5. If the user is satisfied or simply wants to end the retrieval session, he stops. If not, he gives feedback about these results. 6. The engine uses this information and tries to find the most relevant results, and then moves to Step 4. Human beings are at the centre of any image retrieval method since it is primarily their needs that the retrieval engine must cater to. In this way, the person who uses the services of a retrieval engine is in continuous interaction with it, and, at different stages: creating the query, examining the results, evaluating the engine, etc. The objective of this chapter is to provide an overview of the different steps during which the user interacts with the machine in the context of image retrieval. We should point out that this chapter is in no way a survey of existing image retrieval engines and retrieval techniques. The user interested by this type of survey can find a lot of good articles in the literature. For example, [1][2][3][4] and [5]. The chapter has been organized as follows: Section 2 discusses the different types of interactions between human and the engine, whereas Section 3 explores the different tools at the user's disposal for creating his query and the manner in which this query can be created. In Section 4, we will discuss similarity measures and their link with human judgement, and in Section 5, will focus on relevance feedback. In Section 6, we will delve into more detail about the different methods of viewing the results, and Section 7 covers engine performance evaluation. We will end the chapter with a short conclusion. The user expresses his needs using text: The first image retrieval engines used the same query formulation technique as the previous text retrieval engines. This technique involves allowing the user to provide a textual description of what he is looking for. The textual description can be either a group of keywords or a sentence. Keyword queries: The user expresses his needs by providing a keyword, such as in the following query: I am looking for an image that contains "an apple". Most engines enable the user to provide www.intechopen.com
doi:10.5772/5929 fatcat:ij2qro64k5fsnfpl22fibz3lku