CLOSED-LOOP CONTROL OF SEGMENTED IMAGE QUALITY FOR IMPROVEMENT OF DIGITAL IMAGE PROCESSING UDC 62-52:004.932
Danijela Ristić–Durrant, Axel Gräser · 2008
Abstract. In this paper the novel idea of including feedback control at the image segmentation level to overcome the problems of traditional digital image processing is presented. The main idea behind this is that the introduced closed-loop drives the current quality of the binary-segmented image to the desired quality, in order to provide the reliable binary image input to the higher processing levels. The conditions that a measure of image quality should satisfy to be used as the controlled variable in a closed-loop are discussed. The two-dimensional entropy of segmented pixels is proposed as a novel measure of the quality of a binary-segmented image that can be used as the feedback variable in the closed-loop control of image segmentation. Besides the input-output selection, both the selection of control structure and feedback control design are considered. The emphasis is on the specific features of image processing that make closed-loop control in this technology different from conventional industrial control. The benefit of feedback control in image processing is illustrated by its successful application to two real-word applications: object recognition in service robotics and corner detection in industrial images for ship welding. Key words: feedback control in image processing, image segmentation, industrial and service robotics image processing applications 1.