Resolution is commonly used to describe the row and column count of a CCD chip. Essentially, a CCD chip serves as a sampling device, whose maximum sampling rate is governed by the sampling theorem. The sampling rate must be more than twice the Nyquist frequency.

The sampling theory is widely applied to one-dimensional time signals, yet not fully adopted in signal sampling of CCD chips. The sampling rate of CCD chips can be improved via sub-pixel arithmetic. The core principle regards a single pixel as a sub-image composed of multiple sub-pixels. Images with a sub-resolution of 10×10 sub-pixels are generally processable. Calculating the centroid of a spot is a typical application. Due to integration characteristics, the position error of original pixels equals their output value.

Take a one-dimensional grayscale image as an example. If grayscale inflection points fall exactly on pixel edges, contour positions can be pinpointed accurately. In practical scenarios, inflection points seldom appear at ideal positions, making it impossible to precisely locate where inflection points cut through pixels on the chip.

Furthermore, blurred grayscale produces tonal differences, enabling sub-pixel positioning based on grayscale variation. Accurate sub-pixel algorithms can only be achieved when analog images captured by the CCD chip are restored to the greatest precision in the memory of the image processing unit.