8

Application of Attention Models in Image Processing

Visual signals are widely used in daily life since they can faithfully and effectually represent the environment and events around us. There is a saying: a picture is worth a thousand words. Therefore, cameras prevail in consumer electronics such as mobile phones and notebooks. A complete image processing chain includes many different stages, such as signal acquisition, denoising, enhancement, transformation, compression, transmission, reconstruction and display. When an image is processed, visual attention should be considered in several or all of the processing stages to make the system more orientated towards the human vision system (HVS). The application of computational attention models in image processing can also bring about many other advantages such as performance improvement, resource saving and cost reduction. Some stages/areas are selected in this chapter as examples to demonstrate in detail how and why attention models are used in image processing. Methods used in these examples can be easily extended to the other image processing applications.

As discussed in the previous chapters, visual attention models are either pure bottom-up or a combination of bottom-up and top-down, depending on the application under discussion. The procedures for integrating the visual attention model are largely the same for both model types as will be presented in the detailed description of the applications in the following sections.