Feature detection algorithm based on a visual system model

Abstract

An algorithm for the detection of visually relevant luminance features is presented. The algorithm is motivated and directed by current models of the visual system. The algorithm detects edges (sharp luminance transitions) and narrow bars (luminance cusps) and marks them with the proper polarity. The image is first bandpass filtered with oriented filters at a number of scales an octave apart. The suprathreshold image contrast details at each scale are then identified and are compared across scales to find locations in which the signal polarity (sign) is identical at all scales, representing a minimal level of phase congruence across scales. These locations maintain the polarity of the bandpass-filtered image. The result is a polarity-preserving features map representing the edges with pairs of light and dark lines or curves on corresponding sides of the contour. Similarly, bar features are detected and represented with single curves of the proper polarity. The algorithm is implemented without free (fitted) parameters. All parameters are directly derived from visual models and from measurements on human observers. The algorithm is shown to be robust with respect to variations in filter parameters and requires no use of quadrature filters or Hilbert transforms. The possible utility of such an algorithm within the visual system and in computer vision applications is discussed.

The effect of eccentricity and the adapting level on the café wall illusion

“…This suggests that each element of the quadrature pair could be a different mechanism. In addition, Peli (2002) demonstrated that an in-phase (sinusoidal) Gabor filter is sufficient for detecting both bar and edge features in natural images. Those studies cast doubt on the biological plausibility of the idea of a strict quadrature relationship between adjacent simple cells in V1.…”

“…How the information obtained from multiscale spatial filtering is combined is a challenging problem, and to date, many different algorithms have been proposed (e.g., Freeman & Adelson, 1991;Georgeson & Meese, 1997;Marr, 1982;Marr & Hildreth, 1980;Morrone & Burr, 1988;Peli, 2002;Watt, 1990;Watt & Morgan, 1985).…”

Visible Edges Thresholding: a HVS based Approach

“…In this aim, statistical approaches have been developed [13]. Otherwise, approaches inspired by the human visual system (HVS) characteristics have been tested [1,10]. Actually, our method belongs to this family.…”

Perceptual Hysteresis Thresholding: Towards Driver Visibility Descriptors

“…Otherwise, approaches inspired by the HVS characteristics have been tested, e.g. [27,15]. In our application context, it is straightforward to understand that our proposal belongs to the latter family.…”

“…Using this VL information, a hysteresis thresholding is applied using predefined low and high thresholds on VLs and is used to select visible edges in natural images from the precomputed edge map. Compared to [15], we do not wish to automatically extract all visible edges and do not mimic the biological structure of the HVS like in [27]. Our aim is rather to select the edges in the image according to their perceptual relevance.…”