Hi ,
I am Muhammad Walid. need to get this code to understand how wavelet transforms are able to detect moving objects?
I'm new in this field, i want access to learn. thank you
Object recognition is a process of detecting and recognizing certain kinds of objects such as chairs, guitars, buildings, etc., from an image or video sequence. Several investigations have been made so far to extract the object characteristics of real-world images using various approaches ranging from appearance-based approaches such as PCA, LDA, ICA, invariant moment, shape context, SIFT, etc. We use a well-known model based on the "Gabor wavelet transform" approach to extract the characteristics. Gabor wavelets exhibit desirable characteristics of spatial location and selectivity of orientation. It has several advantages against sturdiness, lighting, multi-resolution, and multi-orientation. Object classifications are important areas in a variety of fields, such as pattern recognition, artificial intelligence, and vision analysis. Therefore, the use of the classification of the Gabor wavelet characteristics is performed by several well-known classifiers such as KNN, Neural Network (NN), SVM and Naive Bayes classifiers.
When discriminating features are extracted in optimized locations using selected Gabor wavelets, classifications are made through SVM. Compared to the conventional Gabor function-based object recognition system, the system developed in this paper is robust and efficient. The proposed framework has been successfully applied to two object recognition applications, ie object / non-object classification and face recognition.
The remarkable property of Gabor's multi-scale mapping of features is encountered with scale-space approaches, that an original Gabor-filtered image with individual frequency levels approaches the correspondingly smoothed sub-sampled image with the step filter low. The multi-scaled feature mapping is used to effectively reduce computational costs in filtering. In particular, we show that the Gabor multi-scaled mapping plays an effective role in the match between an input image and the representation of the model for the detection of objects.