A method for extracting 3D‐element based on local cost metrics resistant to radiometric distortion for realistic media application

Abstract

AbstractThe intensity values of pairs of stereoscopic images could be affected by radiometric elements such as camera exposure, illumination direction, etc. Therefore, it is a challenging problem to get consistent and exact disparity values under a variety of real‐world environment. Here, the proposed approach extracts disparity values based on a local cost metric and is resistant to global and local radiometric changes. It is designed in a way that disparity values are extracted using local uniformity of pixel values between pairs of stereoscopic images. The proposed metric‐based method is defined as the resistant local cost metric (RLCM) model. As a result of the experiment, the proposed model performed better than the comparative methods under experimental conditions in which the radiometric distortion between pairs of stereoscopic images was significantly different. The average peak signal‐to‐noise ratio (PSNR) value of the RLCM method is 20.0 dB under the global radiometric change condition and 19.3 dB under the local condition, showing the highest value compared to the comparison group. Consequentially, it is demonstrated that the proposed method is less sensitive to various radiometric changes.