Figure 8.

Local cross-correlation model for disparity estimation. The upper two panels are the half-images presented to the model. The reader can see the sawtooth corrugation by divergent or cross fusing. A Gaussian correlation window W_L is placed in the left eye’s half-image. That window was moved along a vertical line as indicated by the arrow. For each position of W_L, an identical Gaussian window W_R was placed in the right eye’s image, and the cross-correlation (Equation 9) was computed between the two windowed images. Throughout the manuscript, we refer to the size of W_L and W_R as the “window size.” By this we mean two standard deviations of the Gaussian. The lower panel shows an example of the output of the cross-correlator. The abscissa is the position of W_L along the vertical line in the left eye’s image, and the ordinate is the relative horizontal position of W_R; this corresponds to the horizontal disparity. Correlation is represented by color, red for correlations approaching 1, green for correlations near 0, and blue for correlations between −1 and 0. In this example, the sawtooth is revealed by ridges of high correlation.