Automatic out of Range Photo Recognition Using Digital Techniques

This paper addresses digital techniques used to identify automatically those color photographs whose optical densities fall outside acceptable norms for a visual system. Targets are of two diierent kinds indicating the need for diierent recognition algorithms. First, the investigation focuses on color pictures showing high contrast in optical densities such as pictures taken with ash and pictures taken against the light; it estimates statistical parameters which determine mixture densities. Results of Maximum Likelihood, Moment estimation using Prony's Method, Multi scale Analysis and Marquardt optimization algorithm applied to the histogram recognition problem have been compared in terms of eeciency and speed for both low and more detailed vision. Attention is then focussed on pictures showing an optical density gradient in a given direction. Solutions such as plane tting, robust Median Based Etimator are compared with a new " 4-step Median Based Estimator".