Frequently encountered problems on bilevel thresholding of floc images include information loss during the image conversion stage, sampling bias owing to spatially inhomogeneous luminous flux over an optical microscope image or among a series of images, thresholding error adopting various algorithms, and resampling errors during three-dimensional reconstruction. Among these steps the choice of thresholding algorithm could yield most of the processing error. Ways to reduce the potential processing errors are discussed.