Research On Glasses Detection In The Face Image Based On Rectangular Form Filter

Glasses detection plays an important role in face detection and face recognition, with the task of indicating whether there exist glasses in a given image. Accuracy of glasses will directly affect the subsequent face processing system. The results show that the algorithms based on the difference form between the glasses shape and the Background is direct and effective.The form and shape are different concepts, in general, the shape means the geometry in the target binary image, but form refers to a visual form of the target in the grayscale image (with a certain geometry of the gray degree distribution). Rectangular shape and rectangular form are both the most basic elements and they have important applications in image analysis.This thesis proposes a novel glasses detection algorithm in the face image based on rectangular form filter on the basis of previous methods. Firstly, a rectangular form filter algorithm is proposed, and algorithm designing, the calculation process, and features of the algorithm are the focus of this paper. The theory of the rectangular filter form and the characteristics of structural elements are explained. The results of rectangular filter form will change with the different parameters of structural elements and the different modes of filtering. Then in this paper, the automatic glasses detection algorithm based on the rectangular form filter is proposed, which consists of two parts—glasses feature region extraction, detection of rectangular form. In the algorithm of glasses'extraction, firstly the face candidate region are detected, and then we detect the beam of glasses based on the scale of the face region. The second part of the glasses detection is to detect glasses based on the rectangular form filter. In order to eliminate background noise, gray-scale image weighted distance changes is used to preprocess the input images, and then the region of rectangular shape will be detected by the rectangular form filter,and finally, glasses can be identified based on form similarity algorithm. The main contributions of the thesis are summarized in the following.(1)The concept of a rectangular form filter is proposed based on mathematical morphology and this paper give the implementation of RFF.(2) The concept of form similarity is defined and the calculation method is proposed.(3) The rectangular form filter is applied in glasses detection of the face images, which is used for the detection of rectangular form in the natural gray-scale image.This thesis implements the application of the rectangular form filter glasses in the face image to detect glasses. The experimental results demonstrate that the proposed algorithm is capable of achieving accuracy performance, and is an effective solution to the issue of glasses detection in face image.