An image-processing online marking system replaces manual grading for small/medium exams via synchronization alignment, tilt correction (Canny/Hough), adaptive barcode recognition, and Otsu-based objective/subjective analysis.Firstly, in image preprocessing, for the answer card with synchronization head, first find the location of the synchronization head, then use Canny operator to detect the edge of the answer card, and then analyze the important parameter - inclination angle through Hough transform. Secondly, one-dimensional bar code is used to identify student information. When the bar code is deformed, the position of one-dimensional code can still be found accurately, and the student information represented by the bar code can be identified. The efficiency is very high. Then, the objective questions are answered in the form of answer cards. Every five questions are divided into an area, and the question number is written on the top or left, which is clear at a glance. Firstly, the overall area of the objective question is determined according to a certain size, and then it is evenly divided into rectangular blocks of the same size as the answer rectangular box, and then the Otsu method is used to calculate the self adaptation threshold. Finally, the answer sheet of the subjective question uses the horizontal projection method to accumulate the pixel values of the digital image line by line. The system can improve the efficiency of marking, shorten the time of marking, enhance the confidentiality of the test paper, and better safeguard the interests of candidates.