Fig. 10
From: Extended families of 2D arrays with near optimal auto and low cross-correlation
a Output screen from wmAvs before compression. Left—(i) correlation as a 1D stream. (ii) Activity graph of energy difference between adjacent frames and (iii)—merit factor for peaks in a 2D slice correlation. (iv) Sorted correlation peaks and noise. Right—(i) peak without Laplace filter. (ii) After Laplace filter. b Output screen after H264 compression. The sign of the correlation peak in (i) is due to a sync error. Slight noise increase in (iii) is due to compression loss. In (iv), the peak has split due to sync loss causing the correlation to straddle slices. c Left—(i) 136 × 136 marked image with each pixel replicated eight times in x,y to reduce compression damage. (ii) Cross-correlation with the template as an image. (iii) Laplacian of (ii). Right—Latent image is first reduced by 8 × 8 with, each block replaced by its average value. d A watermarked frame. The forehead is saturated, and the background is 0. The watermark is not added to those pixels