The problem tackled by this paper is how a combination of features can improve classification on datasets of similar classes. The work with a flower data set that has large between class similarity and small within class similarity.
Before they use the features they segment the flower in the image. They use the process described in [1]. A thing to notice here is that they were able to segment elongated flowers that do not follow the general flower model described in [1]. This is something good as my understanding after reading [1] was that the segmentation method would probably not work for elongated flowers.
The paper uses 4 types of features to describe the images:
- Colour: It’s based on HSV colour space and k-means clustering. The thought here is that the colour is a feature that can be used to differentiate between certain types of flowers.
- SIFT on foreground region: They refer to the foreground feature as the part of the image which is inside the flower boundary. These features are to gather information about the texture and shape of the flowers.
- SIFT on the foreground boundary: It’s basically the same process as 2 but done in the boundary of the flower.
- Histogram of Gradients: This is used to extract the arrangement of the global distribution of the plant parts. They calculate the HOG features within the smallest bounding box of the segmented flower.
After calulating the 4 features they are combined in an SVM classifier. The kernel is a sum of the individual feature kernels.
Among other things, the results show how the combination of features increase the classification accuracy. It compares classification with one of the features only versus the combination of the four.
For our purposes I think this methodology is the way to go. A combination of the most relevant features is something that should result in a nice classification.
[1] delving into the whorl of flower segmentation
Joel's Blog 