Top-Down Biasing and Modulation for Object-Based Visual Attention
| Autor: | Liang Zhao, Marcos G. Quiles, Alcides X. Benicasa, Roseli A. F. Romero |
|---|---|
| Rok vydání: | 2013 |
| Předmět: |
Visual search
Ground truth Computer science business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Cognitive neuroscience of visual object recognition Image segmentation Top-down and bottom-up design Fixation (visual) Visual attention Computer vision Artificial intelligence business Object-based attention |
| Zdroj: | Neural Information Processing ISBN: 9783642420504 ICONIP (3) |
| DOI: | 10.1007/978-3-642-42051-1_41 |
| Popis: | This work presents a new object-based visual attention model with bottom-up and top-down features. Bottom-up attention is related to the contrast of primitive visual features, such as color, orientation, and intensity. On the other hand, top-down attention is related to the intentions of the viewer and can be seen as a modulation process through the selection system. Thus, if the viewer is searching for an specific shape or color, the top-down modulation can bias the searching process in relation to those features. Our model is composed of five main modules which are responsible for the extraction of the visual features, image segmentation, object recognition, object-saliency map, and object selection. Results on natural images are compared with state-of-the-art approaches and an ground truth fixation maps for a variety of images revealing the efficacy of the proposed approach for visual attention. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|