FIGURE I - uploaded by Truc Hung Ngo
Content may be subject to copyright.
AN EXAMPLE OF WAVELET DECOMPOSITION. (A) RESULTS OF DECOMPOSITION AT LEVEL 1. (B) RESULTS OF DECOMPOSITION AT LEVEL 2. IN THE FIGURE, THE VALUES OF APPROXIMATIONS LL ARE SCALED INTO IMAGE SCALE FOR DISPLAY.
Source publication
Facial paralysis is a common clinical condition occurring in the rate 20 to 25 patients per 100,000 people per year. An objective quantitative tool to support medical diagnostics is necessary. This paper proposes a robust method that decomposes the images into multi frequencies-space domain, and then features are extracted for classification using...
Contexts in source publication
Context 1
... band (LH 1 ), the vertical detail sub band (HL 1 ), and the diagonal detail sub band (HH 1 ). Then the approximation sub band LL 1 in the basic level is used for decomposition in the second level; the approximation LL 2 in the second level is used for the next level decomposition and so on. The dimensions of sub band equals a half of the input. Fig. 1 shows an example of wavelet decomposition into 2 levels of an input ...
Context 2
... block diagram of our system is shown in fig. 2. The first frame of each expression is used as the reference frame for the structured construction of local regions. The face is International Conference on Computer Information Systems and Industrial Applications (CISIA 2015) firstly detected, normalized by the inter-pupil distance, and rotated so that the inter-pupil line is made ...
Context 3
... so that the inter-pupil line is made perpendicular to the vertical face midline. In the second step, the local regions or the regions of interest (ROIs) of the face such as eyebrow, nasal, and mouth regions are detected. Each ROI is divided into 2 equal regions, one on the left and one on the right side, vertically symmetric each other as in fig. 3. Then, each ROI of each frame is normalized in intensity, transformed into scale-space domain using wavelet decomposition. Next, the features are calculated based on the coefficients of the wavelet decomposition and are used as the inputs of an SVM for training and testing. ...
Context 4
... corr stands for correlation coefficient. Fig. 4 shows the calculation performance of the asymmetric feature between the B−R and the B−L regions at decomposition of level 1. 2) Motion feature. The motion feature at a level of decomposition is measured by the correlation coefficient between an ROI in frame−1 and itself in another frame. For example, the motion feature of B−L region of ...
Context 5
... decomposition method is better results than the other methods. Compare with the best cases of the other methods, respectively, our method improves the recognition rates 8.26%, 6.9%, and 4.55%; and reduces average score errors 0.27 pts, 0.02 pts, and 0.09 pts for the raise of eyebrows, the closure of eyes gently, and the nose screw-up expressions. Fig. 5 and fig. 6 present the graphic comparisons of the recognition rates and the average score errors, respectively. They show clearly that our proposed method is better than the other ones. ...
Similar publications
3D facial expression recognition has gained more and more interests from affective computing society due to issues such as pose variations and illumination changes caused by 2D imaging having been eliminated. There are many applications that can benefit from this research, such as medical applications involving the detection of pain and psychologic...
Facial paralysis is a popular clinical condition occurring in 30 to 40 patients per 100,000 people per year. A quantitative tool to support medical diagnostics is necessary. This paper proposes a simple, visual and robust method that can objectively measure the degree of the facial paralysis by the use of spatiotemporal features. The main contribut...
This study proposes a robust method for the face recognition from low-resolution red, green, and blue-depth (RGB-D) cameras acquired images which have a wide range of variations in head pose, illumination, facial expression, and occlusion in some cases. The local binary pattern (LBP) of the RGB-D images with the suitable feature dimension of Depth...
Citations
... The use of LBP provides a better tolerance against variation of illumination, but noise and redundant frequencies still have not been addressed. In previous researches, we proposed a combining LBP with Gabor filters [3], multi-resolution analysis [4]. Using frequency techniques easily removes unnecessary frequencies. ...
... However, a combination of these frequency techniques is not mentioned for a complete system. In addition, the reference [4] did not solve problem of variation of illumination. This paper presents a technique which uses LBP transform to be against variation of illumination; then LBP images are filtered by Gabor filters and wavelet decomposition for feature extraction. ...
... Table 2 presents the recognition rates of using of LBP images filtered by Gabor filters (GBLBP), LBP images decomposed by wavelet transform (WLLBP). Our results are also compared with other conventional methods such as method using intensity of pixel (IP) [1], [3], LBP images (LBP) [2], [3], [7], intensity image filtered by Gabor filters (GBIP) [3] and wavelet decomposition of intensity images (WLIP) [4]. From the result table, we can see that the average recognition rates of GBLBP and WLLPB are better than the conventional methods. ...
