Abstract: A depth-of-field element theory is presented to conquer the limitation that the 3D depth-of-field image only has the distance information.The normal vector clustering method based on depth-of-field element is applied to recognizing depth-of-field image.Though depth-of-field element,Sobel operator is introduced into the detection of depth-of-field image, and a simple edge detection way is proposed for depth-of-field image.An embedded system using ARM(advanced RISC machine) + DSP(digital signal processing) is adopted as the platform for data processing and visualization to meet the requirement of 3D depth-of-field image reconstruction and recognition for real time performance.Experimental results validate the effectiveness of the 3D recognition and edge detection method based on depth-of-field element.