Abstract: This paper proposes an effective approach to extract straight-line feature from digital image. This approach is implemented in a collection of edge detecting based on discontinuity and continuity variation in image and straight-line estimation based on least square template matching algorithm.First, the convolution by LOG algorithm is done to the original digital image to locate the edge by detecting discontinuity variation in image.Then we apply the contour tracing principle to edge fellowing process to find out edge contour and sequentially get initial line feature by splitting the edge contour into short and straight line segment with a rigid threshold.Last, the least square template matching algorithm is used to estimate the initial line by considering it as a whole space for all edges belonging to it and to make a geometrical correction only if it really exists.