Histogram memory reduction in FPGA gradient edge detectors

Gradient edge detectors create an edge map in two steps. First, the gradient image is calculated using appropriate convolution masks. Then, a binary edge map is obtained by thresholding the gradient image. The threshold value significantly affects the final edge map. The most used threshold determination algorithms, Otsu's and Rosin's methods, are based on a gradient image histogram. In some FPGAs, the size of the histogram memory can be a deciding factor in whether an edge detector can be implemented. Based on the analysis of the images of the three databases, we proposed the concept of gradient histogram memory, which reduces the memory size by thirty percent.