2.3.4. Spatial Masking

    When random noise of a uniform level is added to an image, it is much more visible in a uniform background region than in a region with high contrast. This effect is much more pronounced than the effect of the brightness level on noise visibility discussed in Section 2.3.1. Consider the image in Figure 2.18. which illustrates the effect of overall brightness level on noise visibility. In the figure. the noise is much less visible in the edge regions than in the uniform background regions. In addition, the noise in the dark edge regions is less visible than the noise in the bright uniform background regions. One way to explain this is to consider the local signal-to-noise ratio (SNR). If we consider the local SNR to be the ratio of the signal variance to the noise variance in a local region, at the same level of noise, the SNR is higher in the high-contrast region than in the uniform background. Another related view is spatial masking. In a high-contrast region, signal level is high and tends to mask the noise more.

The spatial masking effect can be exploited in image processing. For example, attempting to reduce background noise by spatial filtering typically involves some level of image blurring. In high-contrast regions. where the effect of blurring due to spatial filtering is more likely to be pronounced. the noise is not as visible. so little spatial filtering may be needed.