Linear feature detection is very important in computer vision, image segmentation and pattern recognition. The drawbacks of the standard Hough transform (SHT) are the quantization error and the location of line-segments. In this paper, the nonuniform quantization of HT parameter is proposed to decrease the influence the uniform quantization error of line-segments detection. Moreover, transforming HT parameter to relative parameter and decomposing the digital line to different scale line segments an increase the detection veracity. Experimental results are included to show that the proposed method can achieve high accuracy of line-segments detection and has robustness in the presence of noise.