Multi-pose Face Detection Based on Adaptive Skin Color and Structure Model

Man machine interface by video analysis becomes popular recently. The most frequently adopted body parts for computer interaction are face and hands. Therefore, it is a very important topic to accurately extract face and hand regions from a sequence of images in real time. In this paper, we propose an adaptive skin color model which is based on face detection. Skin colors were sampled from extracted face region where non-skin color pixels like eyebrow or eyeglasses could be excluded. Gaussian distributions of normalized RGB were then used to define the skin color model for the detected person. To demonstrate the robustness of proposed model, experiments under diversified lighting and background were tested. Traditional methods based on RGB, Normalized RGB, and YCbCr were all implemented for comparison. From experimental results, skin color pixels could be detected for each person. The accuracy rate is 95.73% on average and is superior to previously mentioned methods.

A method for the detection and tracking of human face in color images is described in this paper. A skin color model in r,g,b chrominance space is used for segmenting skin color regions from non-skin color region. The best-fit rectangle of the skin color region is labeled as a candidate face. But the face skin color is sensitive to the change of the environment illumination. An algorithm is proposed to update the skin color model's parameters in time so that the model is adapted to different lighting conditions. The confidence measure is presented to evaluate the reliability of skin color model. Experiments demonstrate that the self-adaptive color model is more effective than the fixed model. The color adaptation makes that the color model can be better fit to the more complex application environment.

To realize the eye pupil center location in color face image, a new algorithm based on skin color segmentation and radial symmetry transform is proposed. The algorithm makes full use of the pupil and the surrounding region’s gradient variation characteristics and radial symmetry characteristic, which can effectively solve the interference factors such as uneven spot, frame. Step: First, use the skin color model to detect the face region, and then remove the small block of the background with similar skin color by morphological opening operation and area comparison filtering method. Then, use the horizontal integral projection to determine the candidate eye regions. Finally, apply the radial symmetry transform to the candidate eye region to locate pupil center. Experimental results show that the proposed algorithm has good localization effect, and has strong robustness to various disturbances. Introduction In recent years, line of sight tracking technology is an important and active research area of computer vision. It is widely used in human-computer interaction, auxiliary driving vehicles, virtual reality and military fields [1]. An important process in the line of sight tracking system is the extraction and detection of line of sight characteristic parameters. Eyes as the most important visual organ contains a wealth of sight feature parameters such as the pupil center. Therefore, a fast, accurate and effective eye location algorithm is very important. Proposed Method Fig.1 shows the pupil center location algorithm flow. First, detect face region by the skin color model in the YCbCr color space. Then, using gray level features of eye region, the candidate eye region is detected by the horizontal integral gray projection algorithm [2, 3, 4]. Finally, apply the radial symmetry transform to the candidate eye region to locate pupil center. Face Detection Based on Skin Color Segmentation. Skin color is one of the most striking features of color face images, which is relatively concentrative and stable, and not affected by the brightness of the image. Skin color has a strong clustering feature. Use the ellipse skin color model proposed by Hsu[5] based on YCbCr color space, Where the Y represents the brightness information, and Cb and Cr represent the blue chrominance component the red chrominance component respectively. Ellipse skin color model is shown in Eq.1 and Eq.2, Where Cband Cr represent the original blue component and red component in the YCbCr color space respectively, and x and y represent chrominance components after transformation. Bring X, Y into the left of 4th International Conference on Computer, Mechatronics, Control and Electronic Engineering (ICCMCEE 2015) © 2015. The authors Published by Atlantis Press 1040

Automatic face detection in colored images is closely related to face recognition systems, as a preliminary critical required step, where it is necessary to search for the precise face location. We propose a reliable approach for skin color segmentation to detect human face in colored images under unconstrained scene conditions that overcoming the sensitivity to the variation in face size, pose, location, lighting conditions, and complex background. Our approach is based on building multi skin color clustering models using HSV color space, multi-level segmentation, and rule-based classifier. We proposed to use four skin color clustering models instead of single skin clustering model, namely: standard-skin model, shadow-skin model, light-skin model, high-red-skin model. We made an independent skin color clustering models by converting 3-D color space to 2-D without losing color information in order to find the classification boundaries for each skin color pattern class in 2-D. Once we find the classification boundaries, we process the input image with the first-level skin-color segmentation to produce four layers; each layer reflecting its skin-color clustering model. Then an iterative rule-based region grow is performed to create one solid region of interest which is presumed to be a face candidate region that will be passed to the second-level segmentation. In this approach we combine pixel-based segmentation and region-based segmentation using the four skin layers. We also propose skin-color correction (skin lighting) at shadow-skin layer to improve detection rate. In the second-level segmentation we use gray scale to segment the face candidate region into the most significant features using thresholding. Next step is to compute the X-Y-reliefs to locate the accurate position of facial features in each face candidate region and match it with our geometrical knowledge in order to classify the face candidate region to a face or non-face region. We present experimental results of our implementation and demonstrate the feasibility of our approach to be general purpose skin color segmentation for face detection problem.

This paper proposes a novel technique for detecting faces in color images using AdaBoost algorithm combined with skin color segmentation. First,skin color model in the YCbCr chrominance space is built to segment the non-skin-color pixels from the image. Then, mathematical morphological operators are used to remove noisy regions and fill holes in the skin-color region, so we can extract candidate human face regions. Finally, these face candidates are scanned by cascade classifier based on AdaBoost for more accurate face detection. This system detects human face in different scales, various poses, different expressions, lighting conditions, and orientation. Experimental results show the proposed system obtains competitive results and improves detection performance substantially.

Hand detections based on two proposed pixel-based skin-color models are proposed. Different from the conventional methods, the proposed models are built on the invariant surface obtained by supervised learning on nonlinear color space which effectively preserves vital properties of skin colors in low illumination. By using a high-dimensional feature space, which incorporating probability of fitting the invariant surface with a nonlinear mapping value of chrominance in linear space, first skin-color model is proposed as a thresholding method according to feature values of pixels. Subsequently, a second skin- color model is constructed based on a parametric region which maps the learning data on an invariant surface. This mode achieves more favorable skin-color segmentation results than the conventional methods in experiments.

To detect the color face within a complex background, the paper introduced a face detection method using the skin color feature, the hole feature and the template feature. Firstly, the skin color model has been built in the rgb space, and skin color regions are preliminarily segmented according to the skin color feature; then, the hole feature has been used to realize screening of facial regions; finally, facial regions are matched with facial templates to realize face detection. Experimental results using MATLAB show that the method has good detection effect for images that contain one face or more faces. Thus, it has broad application prospects.

We describe a method for finding the eyes and mouth in a color image of a human face. This method makes use of two new models: a skin color model and a three-dimensional (3-D) geometric model of facial features. The skin color model is based on a medical model for the reflectance of skin and applies to all kinds of skin color. The 3-D geometry model is based on medical statistics of dimensions of the human face and is used to evaluate the geometry of facial features. This model applies for any orientation of the face. The algorithm first determines candidates of facial features by color segmentation. Sets of mouth candidates and eye candidates are evaluated based on the color of the segment that surrounds them and its geometry using the skin color model and the 3-D geometry model. The features are recognized according to the candidate set with the highest likelihood. In experiments, the algorithm detects the features correctly in a large majority of face images in many orientations with a cluttered background.

Skin color model is an important method in skin color detection. This paper introduced the process of building skin color model based on ellipse clustering and detection to image with this skin color model. Before the skin color model building, we analysed color space and chose lighten-unrelated YCbCr space. Because of the effect of environment when get the image, we compensated illumination with logarithm transformation and filter the right light based on statistic. Also, in order to improve the detecting efficiency, we import an ponderance named Ct to present the number of skin pixels. If Ct lower than a threshold, we look it as non-skin pixels. The experiment proved that this skin color model can detect the skin pixels more effectively and have high robust. Compared to the traditional skin color model such as regional model, single gauss model etc, the algorithmical complexity, detecting rate and error detect rate have evident improvement.

Traditional Adaboost face detection algorithm uses Haar-like features training face classifiers, whose detection error rate is low in the face region. While under the complex background, the classifiers will make wrong detection easily to the background regions with the similar faces gray level distribution, which leads to the error detection rate of traditional Adaboost algorithm is high. As one of the most important features of a face, skin in YCgCr color space has good clustering. We can fast exclude the non-face areas through the skin color model. Therefore, combining with the advantages of the Adaboost algorithm and skin color detection algorithm, this paper proposes Adaboost face detection algorithm method that bases on YCgCr skin color model. Experiments show that, compared with traditional algorithm, the method we proposed has improved significantly in the detection accuracy and errors.

This paper proposes a method of segmentation of hand signal, based on the skin color model in YCbCr space and background subtraction under complex background. This paper discussed the reasonable threshold selection of Cb, Cr in the skin color model and the segmentation selection of skin area combining the background segmentation. Finally, an inequality of hands outline feature is proposed to complete the division processing of the palm part of the area of skin. Experiments show the accurate segmentation of gesture under a complex static background.

We propose a vision-based method that robustly extracts hands from backgrounds irrespective of illumination conditions. Many hand tracking systems build a skin color model before the system runs and then they track hands by using the color model. However, the system is unstable because the pre-defined colors cannot be adapted to various illuminations and human skin colors. To circumvent the problem, we postpone the hand color modeling task until the system runs. The system can verify whether an object is a hand or not, and the verified hand is used for modeling a skin color. The method is effective not only for accurate hand extraction but also for reducing noises of backgrounds because the skin color model is optimized to the user's hand and the current illumination. The experiment result shows that the method dramatically improves accuracies of hand tracking and gesture recognition.

In this paper, we proposed an efficient method to address the problem of color face image segmentation that is based on color information and saliency map. This method consists of three stages. At first, skin colored regions is detected using a Bayesian model of the human skin color. Then, we get a chroma chart that shows likelihoods of skin colors. This chroma chart is further segmented into skin region that satisfy the homogeneity property of the human skin. The third stage, visual attention model are employed to localize the face region according to the saliency map while the bottom-up approach utilizes both the intensity and color features maps from the test image. Experimental evaluation on test shows that the proposed method is capable of segmenting the face area quite effectively，at the same time, our methods shows good performance for subjects in both simple and complex backgrounds, as well as varying illumination conditions and skin color variances.

This paper presented a face detection method for the color image using pulse coupled neural network (PCNN) and skin color model. The color image which is processed well through light compensation is converted from RGB to YCbCr color space, then the skin area are divided into sub-block, and skin color segmentation is made for the image in YCbCr space. Finally, we use PCNN to extract all sub-block ignition time sequence, and calculate various sub-block difference degrees between target face and the tested image, if the difference degree is the smallest, then the target face himself is the same person. Experimental results show that the proposed method has higher accuracy and robustness, can obtain satisfactory detection effect.

This paper proposed a fast facedetection method based on the skin color feature and local binary gradient feature. First, according to the clustering of human skin color in the YCbCr color space, the skin color area is detected in the image. Then, it is coarsely fast judged whether if the face is in the skin color area. Finally, the local binary gradient feature is used to judge face accurately, and the weights of the local binary gradient feature is found by using AdaBoost train algorithm. For improved the efficiency of algorithm, the algorithm is accelerated by using the method of the integral image, the cascade classifier and the search order from big to small. The algorithm was tested by a set with 450 color images which the size is 896 ×592. It can find that the average detection time of new algorithm has reduced about 17.1% compare with the face detection algorithm of Paul Viola. The detection accuracy is similar with algorithm of Paul Viola.