Edge Extraction Using Wavelets

In this experiment, we will implement edge extraction using wavelet transforms. Edge extraction represents a fundamental image processing technique that enables the detection of object boundaries and contour features from digital images. We will systematically compare the performance of different wavelet kernel functions (such as Haar, Daubechies, and Symlets) to evaluate their effectiveness in edge detection tasks. The implementation typically involves applying 2D discrete wavelet transform (DWT) decomposition, followed by thresholding coefficients in high-frequency subbands (HL, LH, HH) to enhance edge information while suppressing noise. Through this practical exploration, we gain deeper insights into wavelet applications in image processing and appreciate their role as a "mathematical microscope" for multiscale analysis. To demonstrate real-world applicability, we use license plate localization as our case study - a critical computer vision task that requires accurate detection and recognition of vehicle license plates. The experiment will specifically examine how wavelet-based edge detection improves localization accuracy by enhancing critical features while maintaining computational efficiency through selective coefficient processing. Key implementation steps include image preprocessing, wavelet decomposition level selection, coefficient thresholding strategies, and morphological operations for edge refinement.