Bias-free procedure for the measurement of the minimum resolvable temperature difference and minimum resolvable contrast

Abstract

MRTD (Minimum Resolvable Temperature Difference) The characterization of electro-optical system performance by means of the standard minimum resolvable temperature difference (MRTD) or the minimum resolvable contrast (MRC) has a number of serious disadvantages. One of the problems is that they depend on the subjective decision criterion of the observer. We present an improved measurement procedure in which the results are free from observer bias. In an adaptive two-alternative forced-choice (2AFC) procedure, both the standard four-bar pattern and a five-bar reference pattern of the same size and contrast are presented consecutively in random order. The observer decides which of the two presentations contains the four-bar pattern. Misjudgments are made if the bars cannot be resolved or are distorted by sampling. The procedure converges to the contrast at which 75% of the observer responses are correct. The reliability of the responses is tested statistically. Curves cut off near the Nyquist frequency, so that it is not necessary to artificially set a frequency limit for sampling array cameras. The procedure enables better and easier measurement, yields more stable results than the standard procedure, and avoids disputes between different measuring teams. The presented procedure is a 'quick fix' solution for some of the problems with the MRTD and MRC, and is recommended as long as bar patterns are used as the stimulus. A new and fundamentally better method to characterize electro-optical system performance, called the triangle orientation discrimination threshold (TOD) was recently proposed by Bijl and Valeton (1998).