Sobottka, K., Pitas, I.: Looking for Faces and Facial Features in Color Images. PRIA: Advances in Mathematical Theory and Applications, Vol. 7, No. 1 (1997)  Home/Search   Document Details and Download   Summary   Related Articles   Check

....from being either quite expensive or not very robust. In the first group, we can find systems that are based on exhaustive multiscale searching using neural networks [4] or eigen decomposition [3] and, usually, color is not used. On the other hand, systems that make use of color features [5] [6], are computationally less expensive but are not very robust and present serious problems under uncontrolled environments. The research described in this paper deals with the problem of human face detection on color images and the closely related problems of face segmentation and facial features ....

....on color features which is intended to work under realistic uncontrolled situations. It has been tested using a database of images acquired from TV and from a webcam, achieving very promising results. The key point in face analysis using color images is to search and describe skinlike regions [6]. We have defined a representation space named HIT (Hue, Intensity and Texture) that allows a simpler detection of skin like regions. A fast connected component labeling algorithm is applied on thresholded HIT images, using adaptative thresholding in order to achieve invariance to clutter, noise ....

[Article contains additional citation context not shown here]

Sobottka, K., Pitas, I.: Looking for Faces and Facial Features in Color Images. PRIA: Advances in Mathematical Theory and Applications, Vol. 7, No. 1 (1997)

....For this purpose, we define a representation space named HIT (Hue, Intensity and Texture) that allows a simpler detection of skin regions, using a connected component labeling algorithm. 2. 1 HIT Maps Color features have already been used for face, hand or skin detection, location and tracking [2, 3, 5, 7]. Human skin color is clustered in certain regions of the color space, as already shown by Yang and Waibel in their work about human skin color [2] Differences between individuals are mainly due to distinct melanin concentration, which causes intensity variations. Two color spaces are the most ....

....in their work about human skin color [2] Differences between individuals are mainly due to distinct melanin concentration, which causes intensity variations. Two color spaces are the most widely used for skin analysis: the chromatic color space, or normalized (r, g) 2, 7] and the HSV space [3, 5]. 604 T 604 R 604 H 604 Figure 1: Region searching in HIT maps: input image; hue channel; texture channel; and skin regions found (those verifying size and shape criteria, in white) Our proposal is to use only the Hue component for color analysis. Using only this channel simplifies the ....

[Article contains additional citation context not shown here]

K. Sobottka, I. Pitas. Looking for Faces and Facial Features in Color Images. PRIA: Advances in Mathematical Theory and Applications, Vol. 7, No. 1, 1997.

....Candidates In a preprocessing step, we enhance dark regions in the interior of facial regions by using morphological operations [9] Then candidates for each of the groups are determined by searching for minima and maxima in the topographic greylevel relief. This can be done by using watersheds [10], or as we shall show here, by directly evaluating the x and y projections of the greylevel relief. We compute the y projection of the topographic greylevel relief by determining the mean greylevel of every row of the face region. Then we smooth the y relief by an average filter of width 3 in ....

....to meet a minimum assessment for each criterion as well as a minimum assessment for the weighted sum of all assessments. After candidates are determined, we cluster them according to their left and right minimum x coordinates. This is done using the unsupervised Min Maxalgorithm for clustering [10]. By this step, we reduce the number of candidates significantly and obtain representative candidates for group 1. In analogy, the search for candidates for group 2 and group 3 is done. Examples of x reliefs containing candidates for group 2 and group 3 are illustrated in Fig. 1c,d. 2.3 Best Face ....

K. Sobottka and I. Pitas. Looking for faces and facial features in color images. Pattern Recognition and Image Analysis: Advances in Mathematical Theory and Applications, Russian Academy of Sciences, 1996.

No context found.

Sobottka, K., Pitas, I.: Looking for Faces and Facial Features in Color Images. PRIA: Advances in Mathematical Theory and Applications, Vol. 7, No. 1 (1997)

No context found.

K. Sobottka and I. Pitas, "Looking for faces and facial features in color images," in Pattern Recognition and Image Analysis: Advances in Mathematical Theory and Applications, vol. 7(1), Russian Academy of Sciences, 1997.

No context found.

K. Sobottka and I. Pitas, "Looking for faces and facial features in color images," Pattern Recognition and Image Analysis : Advanced in Mathematical Theory and Applications, vol. 7, no. 1, 1997.

No context found.

K. Sobottka and I. Pitas. Looking for faces and facial features in color images. Pattern Recognition and Image Analysis: Advances in Mathematical Theory and Applications, Russian Academy of Sciences, 7(1), 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC