Abstract | ||
---|---|---|
As a basic operation, image saliency detection has been widely used in various applications. Many approaches have been proposed to detect salient regions. In this study, we learn to detect salient curves of cartoon images based on composition rules. We formulate the detection problem as a binary labeling task where we separate salient curves from the whole curve structures extracted from the cartoon image. We choose a set of curve features related to composition rules, such as the positional relationship between curves and the center, the diagonals, the symmetries or the thirds of the image. We evaluate their effects on salient curve detection by the information gain ranking filter and find that these features play an important role in images respectively with different composition rules. We also use the random forest classifier to build models and use these models to detect salient curves for new images, and it has a satisfactory result. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1109/ICCSE.2016.7581686 | 2016 11TH INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE & EDUCATION (ICCSE) |
Keywords | Field | DocType |
Salienct Curve Detection, Composition Rule, Random Forest, Information Gain | Feature detection (computer vision),Computer science,Salience (neuroscience),Artificial intelligence,Random forest,Computer vision,Pattern recognition,Ranking,Visualization,Feature extraction,Image resolution,Machine learning,Salient | Conference |
ISSN | Citations | PageRank |
2471-6146 | 0 | 0.34 |
References | Authors | |
0 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Chen Chen | 1 | 11 | 2.54 |
Juncong Lin | 2 | 105 | 20.73 |
Minghong Liao | 3 | 90 | 18.97 |
Guilin Li | 4 | 11 | 6.09 |
Guohua Huang | 5 | 0 | 0.68 |