Effect of quantization noise on visual signal detection in noisy images.

Data inaccuracy due to amplitude quantization in digital imaging systems can be viewed as a form of random noise. The effect of this noise is to reduce the accuracy of decisions based on image data. Experimental results are presented to demonstrate the reduction in human-observer decision accuracy due to quantization noise in addition to white Gaussian noise. There is a significant reduction when the ratio Q/sigma p is greater than unity, where Q is the quantization-step amplitude and sigma p is the standard deviation per pixel of the uncorrelated image noise.