Computer Vision News - April 2016

Project Management Tip Why does it happen that project managers are not always satisfied with previous work they have done? What makes a computer vision project successful or not ? Is every image processing expert competent enough to manage a full project? Why do project managers have to learn the hard way all the do’s and don’ts of their job, instead of receiving proper and effective training . Our CEO Ron Soferman has launched a series of lectures to provide a robust and still simple overview of how to ensure that computer vision projects respect goals, budget and deadlines. Mr. Soferman will share every month one of his project management tips in each of the upcoming Computer Vision News issues. This month we are going to learn about: Validating and Maintaining the Quality of Input COMPUTER VISION NEWS 19 One of the main factors contributing to the success of an image processing project is the quality of input . The developer receives images coming from hardware devices (optical systems or scanners) and uses them as samples to construct an algorithmic solution. But during the development of the system, the field implementation or the maintenance phase, problems of image quality might arise: at this moment developer must pay attention not to become the scapegoat of the project and stand on his/her right to receive high quality images for input all along the lifecycle of the system itself. The person in charge of the algorithms should make proper arrangements to exclude any chance of software taking the blame for unsatisfactory output due to poor input quality. This means that several parameters of the input images must be checked and included in the Acceptance Test of the algorithmic work: the overall quality of the image, its sharpness, the noise level, the accuracy of the focus as well as distortions, aberrations and any deviations from the specifications of the optical systems. Among other parameters which must be checked in the input images are the colour separation and the white balance. Moreover, application engineers in the field make frequent calibration tests, just the way we do on our home printers. This too helps the engineer make sure that the input is good enough to do the job correctly. My recommendation is to build, in parallel with the development of the algorithm, a specific test of the images , in order to provide an objective assessment of the input quality. “ Build a specific test of the images, in order to provide an objective assessment of the input quality ”