The key question of computer vision inspection is the edge localizationudprecision in digital images. The method of image primitive localizationudwith subpixel precision is essentially the interpolation in the significantudzones .udThe sensitivity of an image acquisition system is limited by the quantificationuderror, even an input signal is correctly sampled according ta Shannonudtheorem, there are certain input changes that we could not detect . Thisudmakes a « black zone» and il is difficult to prove a measure precisionudhigher than the dimension of the « black zone » .This paper analyses the limite precision of edge localization and dimensionaludmeasure in digital grey images according to dynamic luminanceuddigitization interval and the system impulse response function with the helpudof « black zone » .udThe noise influence on « black zone » is also studied.
展开▼
