| Summary: | 碩士 === 國立成功大學 === 電機工程學系 === 104 === Face recognition plays an important role in nowadays. It is critical in a wide range of applications such as mug-shot database matching, credit card verification, security system, and scene surveillance. Some adversely external factors probably cause the recognition result incorrect. Changes in lighting conditions, facial expressions and pose, occlusion are the main obstacle for face recognition.
In this thesis, variable illumination conditions are considered. We propose a preprocessing scheme to reduce error rates under uncontrolled illumination conditions. Wavelet Transform (WT) is used to decompose an image into low and high frequency components. To deal with low and high frequency components separately, we can alleviate the affects cause by variable illumination conditions. Principal Components Analysis (PCA) and Linear Discriminant Analysis (LDA) are used to extract facial feature.
|
|---|
