Accurate Emotion Detection of Digital Images Using Bezier Curves

Accurate Emotion Detection of Digital Images Using Bezier Curves C.Karuna Sharma, T.Aswini, A.Vinodhini, V.Selvi Abstract Image capturing and detecting the emotions of face that have unconstrained level of problems facing potential applications. There are many efficient approaches for smile detection that differentiates intensity between pixels with grayscale features for face images. The accuracy for picture with pixels will be significantly faster. Thus in our proposed approach we can detect emotions of human from the image with the help of software. First the image will be taken by skin color segmentation and that detects human skin color thus detects the face. Then it segregates eyes and lips from the face. Afterwards it makes Bezier curves for eyes and lips then it compare those images with that already stored on database for each emotion detecting curve possibilities of human face. Then it finds the nearest Bezier curve from database and provides the result as it is stored in Bezier curve emotion as this image s emotion possibility. Thus by using this method emotions expressed by the human face can be detected easily. Accurate result will be produced even for wide image ranges. There is no constraints block for poses given or illumination in the picture. The deformity of any type of emotion is detected at low computational cost. Index Terms Image Software, Bezier curve, emotion detection I. INTRODUCTION In more image processing method that converts images into digital format that perform some different kinds of operations in order to get an enhanced message for extracting useful information regarding image [1]. There are different types of signal dispensation in image input with video frame and photograph where the output may get associated image characteristics with the image. In usual image processing system the images that treats two dimensional signals for applying processing methods with predefined process. For rapid growth in technologies for applications oriented aspects for business and image processing [2]. They form a core research area within engineering and computer disciplines. The most common facial expressions denoting the emotions of daily life events indicating the pleasures, C.Karuna Sharma, B.Tech(CSE), Manakula Vinayagar Institute of T.Aswini, B.Tech(CSE), Manakula Vinayagar Institute of Technology, Kalitheerthalkuppam, Pondicherry-605107,India. A.Vinothini, B.Tech(CSE), Manakula Vinayagar Institute of V.Selvi, Assistant Professor, Manakula Vinayagar Institute of happiness for satisfaction that estimates the mental status of the person [3]. There are many interactive systems for rating the products for distance learning along with conferencing. The statistical analysis of generalized expressions that is active for facial expressions focused on existing works. The works on posing expressions for dynamic movements on facial muscles that pose large amplitude relationships of spontaneous analysis. That is very difficult to intense images for efficient approach that face pixel images. They adopt weak classifiers for pixel differences that have significantly faster and accurate [4]. There are large amount of facial expression recognition correlates specific smile detection. For discriminating the analysis that correlates vector analysis and extraction of lips and cheeks classification. Image processing for importing the optical scanner with digital photography will analyze and manipulate the image including compression of data and compression of images. They enhance the patterns for spotting them cannot be done with human eyes like taken in satellite photographs. The last stage of output where the result will be altered by the image that reports patterns based on image analysis. Usually image processing will be done for object visualization, for sharp and restore of image, retrieval of images, pattern measurement and for objects image recognition. Analog and Digital image processing techniques that use fundamental tool in visual image processing [5]. The collateral image processing techniques uses image as matrix of square pixels is known as picture elements that is arranged in rows and columns. For grayscale image monitoring of each picture element assign intensity for different ranges of shades of grey color. The possible ranges of picture element values are for 256 grayscales which has 8 bit and 16 million colors. The combination of grayscale images that is combined to form images will have vector graphics and bitmap image formats. They form compressed GIF animated format, very efficient JPEG compressed format and other common formats like TIFF, PS, PSD [6]. There is another important thing in image compression is color this have different color formats like RGB with additive color mixing. The secondary color formats of RGB like cyan, magenta and yellow forms mixing of two primary colors and forms new primitive color. There are many image processing applications like intelligent transportation system, remote sensors, tracking moving objects, based on motion detection or recognition based detection. Aerial surveillance method is the most important usage of defense surveillance mechanism used for compact methods for spatial distribution. 929

II. IMAGE PROCESSING APPLICATIONS There are many imaging techniques for analyzing biomedical images that is used in ultrasound and CT scan. Media devices used for smile detection classification used in digital cameras detects smile shutter and clicks images automatically when it is detected [7]. Parameters like surveillance for defense mechanism that distributes information from all the directions. They formulate interpretation of spatial distribution. In biomedical applications for identification for heart disease and lung disease that also detects digital mammograms will automatically inspect them in virtual system. For incandescent lamps for filaments that examine manufacturing process. For essential surface inspection in automatic detection of edge and fractal analysis used for detecting rolled on surface. For components having fault detection of infra red images analyze images with thermal energy distribution [8]. Current image processing techniques diagnose cancer imaging system, brain imaging, neurology which is cost effective and technical efficient. Recognition of data packets authentication that have number of hops detecting compromised nodes. They recover performance analysis of cloned nodes of a network that quantifies the surface vector generalization [9]. For face recognition technology that secure biometric applications that have security with interaction and dependency traits contain variations in face recognition system. For unique challenge in face recognition for real biometric conclusions with facial scar, hair and cuts recognition using consistent finger print identification system. Security access control application for sensing damages finalizes the updates and securing templates. The deformity detection of face images detection on distributed mechanism simulates high efficiency features for recognition process [10]. Properties for extraction with edge line features for intensity of values handling partial conclusions. Classifying limitations for difficult limitation that have different face orientations and eye extraction covers unique identification process. To extract rigid mouth and lips features for facing problems due to picture of nose also. The reliable approach for developing the limitation stated comparing to early methods that structurally shapes active models. To compare challenging recovery features for invisible situation that has large amount of facial expression recognition. Correlating features for specific smile detection with face representation that provides approaches for high accuracy with limited data factor for developing dimensional systems. They track eight mouth points for neural network that classifies smile detection. For different parameters that range imaging for commercial applications that digital camera for accurate real life situations. Receiving smile detection for interesting commercial applications that have digital camera for automatic measurement exhibits accuracy achievement. Face detection in developing smile measurement that has imaging conditions for personal variables for receiving the digital cameras. III. PIXEL BOOSTING EXPRESSION MODEL Facial expression analysis that extracts efficient features that detects facial images based on exploitation of accurate and reliable detection of points tracking unconstrained scenario. Consideration of appearing sensitive feature point detection that has less sensitive facial expression modeling that detects features for encoding. There are more or less sensitive expression analysis for facial modeling that leads to exploitation for response over traditional expression analysis. For location orientation that proves successful expensive computational process for orientation of gradient feature comparison. They can be computed easily with computational efficiency concerning its limited computational resource [11]. The pixel intensity for efficient relationship obtaining high accuracy for different orientations achieves discrimination adopting effective and linear calculations. Performance for weak classifiers randomizing the distribution samples for iteration based classification of samples. Selection of features oriented examples that precise the intensity of difference comprised of threshold indicating inequality signs [12]. The description on wide range of variations will be like age criteria, gender and facial hair factors normalize the partition for illustrating those sample faces for emotion denomination. The rate will be reported average for detecting training data process. Baseline comparison of different pixel oriented factors applying vectors features for performance oriented facial analysis. There is much raw pixel value with division of more facial pixel ranges that is used for direct intensity based detection. The computational cost will have some smile detectors that extracts test image for performing classification for linear training for classifiers that train more than 3000 face images with classifiers for supporting more than 750 vectors. There are many intensity limited differences for each number of pixels of computed images [13]. The weak classifications based on deciding average for cost optimizing. While simulating the results on standard system the faster features based on different approaches improves classification that requires linear time for number of supporting vectors. It costs most expensive methodology which further normalizes the illuminations in the picture. The failure and success rate detection classifies different kinds of pose variations that investigate impacts on further detection of smile emotions [14]. There are more than 4000 face emotion recognition for images with frontal and side with large poses. It reduces range of poses along with frontal and slight improvements that indicates the small difference variations will be harder. Various illumination detection of smile and other emotions with methods proposing variance in recognition of faces with some normalization methods. They remove with state of art methods to follow illumination and normalization methods with simple and wide technique for illustration will be efficient. Basic normalization of photometric luminance images filters images for estimating them. Normalization set of coefficient correspond lowest frequency invariance for illuminant cosine transformation [15]. For one important property tolerating the monotonic illumination changes the invariance by removing the preprocessing effects. There are 930

many series of counter effects that varies with step by step illumination variation for preserving the highlights having important appearances of virtual behavior. Application of facial images that performs detection of smile with baseline approaches. There are many results for baseline approaches that results are obtained for pixel values with improved frequency for accurate methods that fails in achieving pixel values. The illumination values that features normalization for real life complex image recognition. There are some possible reasons for having complex variations in illuminating the image processing [16]. As superior performance for normalization methods that has some data for existing difficulties in handling complex variations in illuminating the data set for performing best process. There is more accuracy in values for small number of images. In emotion detection smiles and others can be detected by calculating for pose given by smiles with real world scenarios. This kind of features is not suitable for featuring image constraints. Illumination of problems that have variation with different poses that found emotion detection in face images. IV. EMOTION DETECTION BY BEZIER CURVES Different scenario that efficiently approaches the intensity having differences with pixels of grayscale face images that use simple features. The adopted choosing and combination of weak classifiers based on pixel differences that forms strong classifiers with smile detection. There are many approaches behind similarly faster having accurate results. The feature based accuracy detection that supports vector for different pairs of pixels. Varying poses along with illumination that have hard constraints analyzing face examining different normalization methods. The illumination methods for investigating the impact of variation on poses given for converting the binary image that fills RGB color using the exact place for image processing. The software used for detecting human emotions from digital images that takes image by segmenting its skin color. This detects skin color of human face that separates the eyes and lip lines from the face of image. Then the process of drawing Bezier curve for eyes and lips. It is then compared with Bezier curves of eyes and lips to the Bezier curves of eyes and lips that are stored already in the database. From the thousands of images stored for emotional facial curves from the database. In the database there are different types of facial expressions stored and different facial vector emotions are stored as image for Bezier curve emotion that provide database for eyes and lips curve lines with emotion criteria that holds upon example details. This kind of digital images have wide range of accurate applications have curve line expression kind of poses provided. There are no constraints of posing kind of illumination that provides maximum ranges that have less computational cost. As they have lesser emotional types of deformity kind of detection will be detected [17]. There are emotional kinds of largest connected regions that perform skin color segmentation. The probable connected regions will become facing open form of largest regions will be connected. There are more height and width for larger and equal ratio of height and width between faces having different expressions. In face detection of converting binary images from red, green and blue colors that converts the average value for each pixel in average values that replace black pixel with white pixel [18]. By the above method the binary image from RGB colors. From the binary image the forehead will be found by scanning the middle of image that need to find a continuous white pixel that continues after black pixel. To find the maximum width of white pixel that searches vertically for both sides. The small width for half broken scan of reaching till the eyebrow then it will cut the face portion taking it as starting position of forehead. For detection of eyes that converts RGB color for face will be changed to binary face. The highest continuous pixel finds height between different ranges of middle position of two eyes. The height and upper position of two eyebrows for vertical searching from middle to right and left eye will position pixels between eyebrows and eyes. There are eyebrow pixels and eye curves connected with vertically connected for black pixels. For left and right eye pixel lines between eyes and eyebrows started from heights of images with width of lower position. There are lower positioning for two eyes by searching lack pixels that vertically in the middle of lower image that is having starting position. The right side for left eye that searches the black pixel for upper and lower position from the left eye for searching the middle of the RGB colors. For lip detection determining the outer curve line of lips that is determined from forehead and eyes added within the distance of lower height of eye with upper height of box will be contained inside the lip outline curve. For detecting eyes and lips it needs to convert binary image from RGB image that will be searched among binary images. As in Fig.1 inside the outline of lips it includes some nose par also so the skin color segmentation should convert skin pixel to white and other black pixels. Fig.1. Applying Skin Color Fig.2. Finding Big connected region for lips The pixels similar to skin color segmentation will be converted into white pixels where the difference will be lesser than or equal to similar pixel. They will find the distance between low and high average RGB value. For finding similar pixel with greater or equal pixels depends on quality of image where its quality is high and the image quality will be very low. The black region defines lips, nose and some other little parts differ from the skin color. Then it applies big connected region that finds lip in binary image. Then here the Bezier curve on binary lip is found from 931

starting and ending pixel of lip in horizontal by finding two tangent points which is not the lip part. Similar process will be followed for finding upper and lower lips and it use cubic Bezier curves for drawing Bezier curve of lips. Thus two curves will be drawn one for lower and one for upper lip. Fig.3. Applying Bezier lines to lips curve For applying Bezier curves to eyes first the eyebrow should be removed from eye then for continuous search of black pixel then continuous white pixel then it finds binary image of eye box. Then continuous black pixel from box which contains only eye will be found. Fig.4. Retrieving Binary image of Eye Thus the eye box containing eyes and skin color around the box will be similar to find the region of eye. They apply big connect that has region where the eye is the biggest thing which is not similar to skin color. Fig.5. Applying Big Connected Eye region Fig.6. Applying Bezier Curve to eye From the fig above the Bezier curve for eye similar to lip will get the shape of eye. Thus from our database two tables are available one for storing peoples name and the index for four different kinds of emotion stored as position index. It has six different control points in lips Bezier curve that has sic control points for right eye Bezier curve, lips height and width then this learns emotions of people by all these studies. For emotion detection of image first it finds Bezier curve of lips with left and right eye. This personal information will be available in the database where the program matches the emotions nearest height of program that gives nearest emotion as output. Thus the emotion information will not be available in database then the program calculates average height for each emotion in database for people that accordingly takes average height decision. V. CONCLUSION In our proposed paper the emotion detection of images of face captured in real world scenarios with interesting problem in many applications that is because of limited computational resources. The desired features will be computed with ease and efficient along with differences of intensity between pixels in grayscale face images that use simple features. We examine the emotions by drawing Bezier curve lines for lips and eyes that has that has to be matched with the emotion curves already stored in database. Then these facial investigations with respect to different poses will be determined. REFERENCES [1] M. Pantic and L. Rothkrantz, Automatic analysis of facial expressions: The state of art, IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 12, pp. 1424 1445, Dec. 2000. [2] B. Fasel and J. Luettin, Automatic facial expression analysis: A survey, Pattern Recognit., vol. 36, no. 1, pp. 259 275, a Jan. 2003. [3] Z. Zeng, M. Pantic, G. Roisman, and T. Huang, A survey of affect recognition methods: Audio, visual, and spontaneous expressions, IEEE Trans. Pattern Anal. Mach. Intell., vol. 31, no. 1, pp. 39 58,Jan. 2009. [4] T. Kanade, J. Cohn, and Y. Tian, Comprehensive database for facial expression analysis, in Proc. IEEE Int. Conf. Autom. Face Gesture Recog., 2000, pp. 46 53. [5] J. F. Cohn and K. L. Schmidt, The timing of facial motion in posed and spontaneous smiles, Int. J. Wavelets, Multiresolution Inf.Process., vol. 2, pp. 1 12, 2004. [6] Z. Zeng, Y. Fu, G. I. Roisman, Z. Wen, Y. Hu, and T. S. Huang, Spon-taneous emotional facial expression detection, J. Multimedia, vol. 1, no. 5, pp. 1 8, Aug. 2006. [7] M. Bartlett, G. Littlewort, M. Frank, C. Lainscseck, I. Fasel, and J. Movellan, Automatic recognition of facial actions in spontaneous ex-pressions, J. Multimedia, vol. 1, no. 6, pp. 22 35, Sep. 2006. [8] N. Sebe, M. S. Lew, I. Cohen, Y. Sun, T. Gevers, and T. S. Huang, Authentic facial expression analysis, in Proc. IEEE Int. Conf.Autom. Face Gesture Recog., 2004, pp. 517 522. [9] J. F. Cohn, L. I. Reed, Z. Ambadar, J. Xiao, and T. Moriyama, Automatic analysis and recognition of brow actions in spontaneous facial behavior, in Proc. IEEE Int. Conf. Syst., Man, Cybern., 2004, pp. 610 616. [10] Y. Shinohara and N. Otsu, Facial expression recognition using fisher weight maps, in Proc. IEEE Int. Conf. Autom. Face Gesture Recog., 2004, pp. 499 504. [11] A. Ito, X. Wang, M. Suzuki, and S. Makino, Smile and laughter recognition using speech processing and face recognition from conversation video, in Proc. Int. Conf. Cyberworlds, 2005, pp. 437 444. [12] U. Kowalik, T. Aoki, and H. Yasuda, Broaference a next generation multimedia terminal providing direct feedback on audience s satisfac-tion level, in Proc. IFIP TC13 Int. Conf. Human-Comput. Interact. (INTERACT), 2005, pp. 974 977. [13] O. Deniz, M. Castrillon, J. Lorenzo, L. Anton, and G. Bueno, Smile detection for user interfaces, in Proc. Int. Symp. Adv. Vis. Comput., 2008, pp. 602 611. [14] D. Freire, M. Castrillon, and O. Deniz, Novel approach for smile de-tection combining LBP and PCA, in Proc. Int. Conf. Comput.Aided Syst. Theory (EUROCAST), 2009. [15] J. Whitehill, G. Littlewort, I. Fasel, M. Bartlett, and J. Movellan, To-wards practical smile detection, IEEE Trans. Pattern Anal. Mach. In-tell., vol. 31, no. 11, pp. 2106 2111, Nov. 2009. [16] Omron Okao Vision System [Online]. Available: www.omron.com/ r_d/coretech/vision/okao.html 436 IEEE TRANSACTIONS ON IMAGE ROCESSING, VOL. 21, NO. 1, JANUARY 2014 [17] G. Donato, M. Bartlett, J. Hager, P. Ekman, and T. Sejnowski, Classi-fying facial actions, IEEE Trans. Pattern Anal. Mach. Intell., vol. 21, no. 10, pp. 974 989, Oct. 1999. [18] P. Viola and M. Jones, Rapid object detection using a boosted cascade of simple features, in Proc. IEEE Conf. Comput. Vis. Pattern Recog., 2001, pp. 511 518. C.Karuna Sharma Currently she is pursuing B.Tech (CSE) at Manakula Vinayagar Institute of Technology, Kalitheerthalkuppam, T.Aswini Currently she is pursuing B.Tech (CSE) at Manakula Vinayagar Institute of Technology, Kalitheerthalkuppam, 932

A.Vinothini Currently she is pursuing B.Tech (CSE) at Manakula Vinayagar Institute of Technology, Kalitheerthalkuppam, V.Selvi she has finished her M.E (cse) and currently she is working as assistant professor in Manakula Vinayagar Institute of Technology, Kalitheerthalkuppam, Pondicherry-605107, India. Her research areas are Computer networks, information security. 933