Hough transform is widely used for detection of some parameters of a figure or figures in a digital image. One of the bottle neck of Hough transform is its processing speed. It takes so much time for accumulation in the θ-ρ plane that caused by the “time consuming” ρ value calculation.
This paper proposes a new fast Hough transform which employs the table look-up method for ρ value calculation. The ordinary table look-up method in Hough transform, in which the sinusoidal function is calculated beforehand and saved in memories, has had the defect that the size of the look-up table has become so much big although the processing speed has been improved remarkably. The new method solves this problem by introducing the “bit block dividing” principle for the image position variables
x and
y. The actual system can easily be realized by using simple hardware devices on the basis of this principle. The total throughput is improved more by employing pipe line processing.
The proposed principle is described and the comparisons with other ordinary methods are mode in the beginning part of the paper. The examination and the trial structuring are explained on the hardware system in the next. The result of the evaluation test on the trial made hardware is shown and the discussion on it is done in the last part.
抄録全体を表示