A two-layer neural network connectionist model of the Hough transform (HT) is proposed for the detection of lines in image space. The model is implemented in the optical domain via an optical matrix-vector multiplication technique where interconnection weights of the neural network are binary and are adjusted according to the line parameters. The dimension of the weight matrix depends only on the dimensions of the image and parameter space and no further dynamic updatings of weights are required for a particular dimensions of image and parameter space adjustment. The peak strength of a neuron in the parameter or Hough space is detected by a winner-take-all optoelectronic circuit. Experimental results are presented.