Figure 2: Schematic diagram of the operation of the proposed two-channel model. The initial stage of processing, common to both channels, is the removal of the mean luminance of the image. The signal in the first-order channel is then subjected to motion energy analysis. Prior to motion analysis, the signal in the second-order channel is pre-processed by a texture grabber. The texture grabbing operation consists of low-pass spatiotemporal filtering, followed by full-wave rectification, followed by further low-pass spatial filtering. The output of the texture grabber is then subjected to motion energy analysis. The mechanism of motion analysis in the two channels differs only in the scale at which it operates: fine-scale (short-range) in the first-order channel and coarse-scale (long-range) in the second-order channel.