Goal of this study:
The visual system identifies surfaces and uses their properties to help recognize objects. One example of this phenomenon is color transparency: when a surface is seen both in plain view and through a transparent overlay, the visual system still identifies it as a single surface. We think that the conditions to see transparency and shadow are closely related. Our goal is to investigate the application of D’Zmura et al’s Model of transparency perception to image segmentation, more precisely on shadow detection. The actual methods to detect shadow are based on invariant color models. We assume that using a perceptual model will provide better results.
Relations between shadows and transparency:
- Transparent overlay created with translations directed to a lower luminance level (Figure 1). Vector length decreases. 2 X-junctions. Translation formulation : b=a+t.
- No transparency effect with shears except small ones (Figure 2). Vector length decreases. 2 X-junctions. Translation formulation : b=a+t, and b=a-t (second vector).
- Relationship between shadow and filter (Figure 3). 2 X-junctions but different configuration. Filter? Projection of shadow? Depending on vector length?
- Relationship between shadow and filter (Figure 4). 2 X-junctions. Here are shears. Light spot + shadow? Depending on vector length? Perception in 3D?
Summer semester project 2003
Shadow Perception and its Application to Image Segmentation
Christian Cornaz, 6th semester
Mounir Krichane, 6th semester
Swiss National Science Foundation (SNF) under grant number 20-59038.99.