Segmentasi Citra Wajah dengan Implementasi Adaptif Threshold- Integral Image

<p>Segmentasi wajah merupakan bagian penting dalam pengolahan citra digital untuk mengetahui objek wajah dalam citra sebelum dilakukan pendeteksian ekspresi wajah. Adaptif <em>Threshold – Integral Image</em> adalah salah satu teknik segmentasi berbasis <em>pixel-based</em>,<em> </em>yaitu <em>local thresholding</em>. Penelitian ini bertujuan untuk memisahkan objek wajah manusia dan <em>background </em>-nya. Citra wajah yang akan digunakan nanti citra di dalam ruangan (<em>indoor</em>)<em> </em>dan di luar ruangan (<em>outdoor</em>) dengan resolusi gambar 300x400 piksel. Pada penelitian ini juga mencari nilai parameter S (<em>kernel</em>) dan T (<em>threshold</em>) yang terbaik dengan melakukan 16 kali percobaan. Dan didapatkan hasil terbaik, yaitu citra di dalam ruangan (<em>indoor</em>) nilai S=1/2 dan T=50, serta citra di luar ruangan (<em>outdoor</em>) nilai S=1/30 dan T=30. Segmentasi citra wajah dengan menggunakan metode Adaptif <em>Threshold – Integral Image</em> <em>robust</em> (kuat) terhadap intensitas cahaya tinggi dan rendah dengan mengatur nilai parameter S (<em>kernel</em>) dan T (<em>Threshold</em>) maka metode ini mampu memisahkan objek wajah dan <em>background</em> -nya. Dari hasil uji coba <em>threshold</em> menggunakan metode Adaptif <em>Threshold – Integral Image</em> terhadap citra di dalam ruangan (<em>indoor)</em> dan di luar ruangan (<em>outdoor)</em> menghasilkan <em>thresholding</em> yang baik dengan mempertimbangkan nilai parameter S (<em>kernel</em>) dan T (<em>threshold</em>) memberikan hasil dengan tingkat akurasi yang tinggi, yaitu citra di dalam ruangan (<em>indoor</em>) sebesar 96.72%, dan citra di luar ruangan (<em>outdoor</em>) sebesar 93.59%.</p><p> </p><p><em><strong>Abstract</strong></em></p><p><em>Face segmentation is an important in digital image processing to find out the object's face in the image before detecting facial expressions. Adaptive Threshold - Integral Image is a pixel-based segmentation technique, which is local thresholding. This study is intended to split the object of a human face and its background. Face images that will be used later in indoor and outdoor with an image resolution of 300x400 pixels. This study also searched for the best S (kernel) and T (threshold) parameter values by performing 16 experiments. And the best results are obtained, name the image in the room (indoor) the value of S = 1/2 and T = 50, and the image outside the room (outdoor) the value of S = 1/30 and T = 30. Face image segmentation using the Adaptive Threshold - Integral Image robust method of high and low light intensity by setting the S (kernel) and T (Threshold) parameter values, this method is able to split the face object and its background. From the results of the threshold trial using the Adaptive Threshold - Integral Image method for indoor and outdoor images produces a good thresholding by considering the values of the S (kernel) and T (threshold) parameters to give results with a high degree of accuracy, that is indoor images of 96.72%, and outdoor images of 93.59%.<strong></strong></em></p><p><em><strong><br /></strong></em></p>

DEEP LEARNING TECHNOLOGY FOR VIDEOFRAME PROCESSING IN FACE SEGMENTATION ON MOBILE DEVICES

The aim of the research is to reduce the frame processing time for face segmentation on videos on mobile devices using deep learning technologies. The paper analyzes the advantages and disadvantages of existing segmentation methods, as well as their applicability to various tasks. The existing real-time realizations of face segmentation in the most popular mobile applications, which provide the functionality for adding visual effects to videos, were compared. As a result, it was determined that the classical segmentation methods do not have a suitable combination of accuracy and speed, and require manual tuning for a particular task, while the neural network-based segmentation methods determine the deep features automatically and have high accuracy with an acceptable speed. The method based on convolutional neural networks is chosen for use because, in addition to the advantages of other methods based on neural networks, it does not require such a significant amount of computing resources during its execution. A review of existing convolutional neural networks for segmentation was held, based on which the DeepLabV3+ network was chosen as having sufficiently high accuracy and being optimized for work on mobile devices. Modifications were made to the structure of the selected network to match the task of two classes segmentation and to speed up the work on devices with low performance. 8-bit quantization was applied to the values processed by the network for further acceleration. The network was adapted to the task of face segmentation by transfer learning performed on a set of face images from the COCO dataset. Based on the modified and additionally trained segmentation model, a mobile app was created to record video with real-time visual effects, which applies segmentation to separately add effects on two zones - the face (color filters, brightness adjustment, animated effects) and the background (blurring, hiding, replacement with another image). The time of frames processing in the application was tested on mobile devices with different technical characteristics. We analyzed the differences in testing results for segmentation using the obtained model and segmentation using the normalized cuts method. The comparison reveals a decrease of frame processing time on the majority of devices with a slight decrease of segmentation accuracy.