A comparison of contrast measurements in passive autofocus systems for low contrast images

A number of contrast measurements have been investigated and compared in the literature. Each of them exhibits an ideal curve with a well defined peak standing for the best focused image. However, a focused image obtained in low light conditions possesses a small contrast value, which may be easily influenced by noise. In this case, contrast measurements may generate fluctuant curves with many local peaks. This paper presents a comparison among six contrast measurements in passive autofocus systems towards a non-previously researched object of low contrast images. The criterium to evaluate the performance of each measurement is unimodality. And we assess the similarity of the resulting curves with an ideal focus curve which exhibits a single peak and an absence of plateau. Experimental results from six typical image sequences indicate that Tenengrad and CMAN approaches yield the best performance, but it is still necessary to derive a more elaborated method because both methods fail to generate a single sharp peak in some circumstances.