Skip to Main Content
This paper proposes self-quotient ε-filter and presents its application to feature extraction from noisy facial image. Self-quotient filter (SQF) is a simple filter defined as the ratio of the input image and its smoothed versions. It is light invariant, and can clearly extract the outline of the object from the image independent of shadow region. However, when the image includes not only signal but also noise, SQF cannot extract the feature clearly. To solve the problems, we look to ε-filter and design self-quotient ε-filter. By defining self-quotient ε-filter as the ratio of two different ε-filters, we can extract the feature not only from facial images without noise but also facial images with noise. Experimental results show that the proposed method can clearly extract face features such as eyes, nose and mouth from noisy facial images.