基于改进Sobel算子的人脸速写系统设计

Design of Face Sketch System Based on Improved Sobel Operator

人脸速写是边缘检测技术的一个特殊应用。针对当前流行的边缘检测算法难以在人脸速写应用中有效实施的难点,设计了一种以NXP公司MK60DN512VLQ10单片机为核心控制单元,以改进的Sobel算子为边缘检测算法的人脸速写系统。系统采用低功耗CMOS图像传感器OV7620进行人脸信息的采集,采用一种改进的Sobel算子进行人脸轮廓的提取,改进的算子采用45°,135°,225°,315°四个方向模板进行边缘检测,通过模板中权值的再分配,可以使模板对不同方向的边缘进行较好的检测,改善了传统Sobel算子精度较低和边缘较粗糙的缺点,提高了边缘检测的精度和连续性。提取到的边缘通过边缘细化和二值化处理后输出控制信息,可分别控制水平和竖直方向步进电机的转动,较好的描绘出人脸轮廓。结果表明,系统硬件设计精简,改进的Sobel算子可快速准确地提取出人脸轮廓,系统可较好的完成人脸速写。

Face Automatic Sketch（FASK） is a special application in edge detection technology. For the implementation difficulty of the edge detection algorithms in face automatic sketch applications, the FAS system is designed in which MK60DN512VLQ10 microcontroller is employed for core control unit. Sensor OV7620 with low power CMOS image sensor is used in this system for face information acquisition. An improved Sobel operator is proposed for facial contour extraction which uses four directions template in directions 45 °, 135 °, 225 ° and 315 ° for edge detection. Using the redistribution of the weights in the template, the edges in different directions are detected accurately by the direction templates, which can overcome the shortcoming of low precision and roughness wherefore the accuracy and continuity of the edge detection can be improved. The information of extracted edges is outputted after the processing of the edge refinement and binarization, which control the horizontal and vertical stepping motor separately for accurate depiction of face contours. The results show that the improved Sobel operator can extract the contours of the human face quickly and accurately, and the system can finish the sketches better.