A computationally efficient technique for real-time detection of particular-slope edges

Identification of oblique lines of a particular slope is needed for various applications such as motion tracking for smart cameras. Wavelets and gradient-based techniques, such as Sobel and Canny, do not classify edges based on their slopes. The Hough transform (HT) does classify edges based on their slopes but with high computational complexity, even using its most improved versions. This paper presents a computationally efficient technique for detecting edges of a particular slope. The angle of the required edges is converted into pixel increments over rows and columns. Using these two simple parameters, parallel, oblique lines of a particular slope are formed. A first-order, orthonormal Haar low-pass filter (LPF) is used over the formed lines to filter out undesired edges. The hardware architecture of the proposed technique is fully described, including processing time, based on the number of clock cycles, and fixed-point implementation. A line-based memory mechanism was used to minimize the memory requirements to two simple registers. To demonstrate the computational advantage of the proposed technique, it is compared to the Sobel, Canny and HT detectors.