Automated threshold detection for object segmentation in colour image

• Main Authors: Akhtaruzzaman, Md., Shafie, Amir Akramin, Khan, Md. Raisuddin
• Format: Article
• Language: English; English
• Published: Asian Research Publishing Network (ARPN) 2016
• Subjects: TA Engineering (General). Civil engineering (General); TA349 Mechanics of engineering. Applied mechanics
• Online Access: http://irep.iium.edu.my/54857/1/54857_Automated%20threshold%20detection.pdf; http://irep.iium.edu.my/54857/2/54857_Automated%20threshold%20detection_SCOPUS.pdf; http://irep.iium.edu.my/54857/; http://www.arpnjournals.org/jeas/research_papers/rp_2016/jeas_0316_3928.pdf

Object segmentation from single background colour image is important for motion analysis, object tracking, trajectory identification, and human gait analysis. It is a challenging job to extract an object from single background colour image because of the variations of colours and light intensity. Most common solution of the task is the uses of threshold strategy based on trial and error method. As the method is not automated, it is time consuming and sometimes a single threshold value does not work for a series of image frames of video data. In solving this issue, this paper presents an Automated Threshold Detection Algorithm, H(•). The algorithm is applied in segmenting human lower limbs from a series of image frames of human walking. The procedure starts with selection of optimal RGB channel. Then H(•) algorithm is applied for automated threshold detection to convert the image frames into grayscale image. In the next stage, Line Fill (LF) algorithm is applied for smoothing the edges of object and finally background is subtracted to extract the targeted object. Results of the applied procedure show that the algorithm is viable to extract object from single background color image and can be used in human gait analysis. © 2006-2016 Asian Research Publishing Network (ARPN).