Object Recognition with Wavelet-Based Salient Points

In this paper, an efficient method to recognize various objects using wavelet based salient points with the help of Moment features is presented. In the detection of salient points, a salient point detector is presented that extract points where variations occur in the image, whether they are corner-like or not. The detector is based on wavelet transform with full level decomposition to detect global variations as well as local ones. This method provides better retrieval performance when compared with other point detectors. After detecting the salient points, patches are extracted over those points. The patches have the advantage of being robust with respect to occlusion and background clutter in images. Then the features are extracted using Basic Moments method for the detected patches in order to give them to a classifier. Support Vector Machines scale relatively well to high dimensional data. SVM classifier recognizes the objects (positive images) from the background (negative images) and vice-versa. The experimental evaluation of the proposed method is done using the well-known and complex Caltech database with complex images. The results obtained here proved that the proposed method is able to successfully recognize the objects with good recognition rate along with the background using wavelet based salient points with full level decomposition under challenging conditions.