Loading...
Please wait, while we are loading the content...
Similar Documents
Human perception based image representation
| Content Provider | Indraprastha Institute of Information Technology, Delhi |
|---|---|
| Author | Seth, Yash |
| Abstract | Images are represented by certain combination of their features known as its descriptor. The general two types of descriptors are: Trained descriptors and Un-Trained descriptors. Trained descriptors include Attributes, Classemes and ObjectBank [2]. Un-Trained include descriptors such as GIST and Vanilla bag of words.One of the major limitations of existing approaches of image representation is that they do not think or perceive an image the way we humans do. With this motivation we are trying to develop an Image Representation mechanism which models itself around human perception. Here,we try to use trained descriptors to mimic the human perception and come up with a representation that recognizes and groups images the way humans would. Also, we see that these descriptors are in some way better at semantic tasks, such as classi cation, than other trained or un-trained descriptors. The performance of the proposed feature extraction approach is evaluated and compared on existing and proposed databases. |
| File Format | |
| Language | English |
| Access Restriction | Authorized |
| Subject Keyword | Image analysis Computer vision Image representation Image classification |
| Content Type | Text |
| Educational Degree | Bachelor of Technology (B.Tech.) |
| Resource Type | Thesis |
| Subject | Data processing & computer science |
