Abstract: In order to solve problem that most 2D (Two-dimensional) shape representations don′t fit for wavelet transformation processing, a new method of 2D shape representation has been put forward. It stores object borderline pixels coordinate value (xi, yi) into two arrays x = X[i]; y = Y[i] by the sequence of object borderline tracking result, object boundary has been represented by this two array. It changes a 2D image problem into two single dimension arrays′ problem and can be processed by wavelet transformation. Then make full use of the ability of time and frequency localization of wavelet transformation and find characters both in time and frequency. The boundaries leaps can be distinguish from noise. This technique has been applied in workpiece boundary noise reducing. And it shows that noise can be eliminated and boundaries corners are remained at same time.