User:Marigoldx/sandbox

Shape From Shading
Shape from shading is the ability to perceive shape and derive the three-dimensional orientation of surfaces. The human visual system can detect the three-dimensional orientation of surfaces by using variations in image intensity. This is often due to shadows, shading and the perception of light. Shape from shading is one of the most important yet poorly understood aspect of human vision. Findings from VS Ramachandran study in 1988 reveal that shape from shading has computational mechanisms underlying this ability: Perception of shape from shading is a global operation which assumes that there is only one light source illuminating the entire visual image, Three dimensional shapes that are defined by exclusively by shading can provide tokens for the perception of apparent motion,  background can also powerfully influence the perception of three-dimensional shape from shading. Research has shown that people tend to assume the light is above and slightly to the left of the object. (Sun and Perona, 1998) This is how human observers resolve shape from shading illusions by assuming that the light source is coming from above-left.

History of research into shape from shading
First introduced by Horn (1970) who stated Shape from shading is the ability to perceive shape from shading and derive the three-dimensional orientation of surfaces. Research into the dilemma of perceiving shape from shading has progressed and changed direction over the years to focus on a range of topics in perception as well as modern technology.

Early Research
The first psychological research into the perception of shapes from shading focused mostly on differences in light, shapes, and explanations through numerical methods. Early shape from shading studies worked to establish, develop and test hypothesis as shape from shading and perception wasn't well understood. This was due to poor technology at the time as in order to discover show shapes in our environment are processed researchers needed to be able to complete the following: To produce visual displays that show different shapes using specific and controlled variables and record the participants responses accurately. In order for this research to be valid the methods would also need to be repeatable for other researchers to test the hypothesis which would have been difficult to prove with limited access to the level of technology required at the time. It would have been very expensive for the average psychology lab to be able to afford, especially since at the time psychologists did not have a well-established attitude to what solid shape with respect for vision is. (Mingolia & Todd, 1986)[5] Until then, shape from shading research had been considered to fall under a different categories in psychology and being due to cognitive reasons or the intellectual abilities of the participants and in line with visual psychology theories. These factors lead to an overall poor understanding of shape from shading theory.

In a study observers judged the slants and flits of numerous regions within shaded images of different surfaces that varied in shape, orientation, surface reflectance, and direction of illumination. The perceived three-dimensional structure of each surface was calculated from these judgments This disproved popular theories at the time as they were not psychologically valid. (Mingolia & Todd, 1986) As the presence of specular highlights or cast shadows and no impact on performance which contradicted previous research. This study also found that many errors in the observers responses was due to a tendency to perceive surfaces whos axes aligned with the display screen.

Establishing research
Research then progressed into subjects such as levels of processing, biases, and motion detection by the early 2000’s. It was understood at the time that we acquire knowledge about three dimensional objects from how they look, which is also determined by other factors such as their placement in comparison to other objects, illumination and processing the image in the brain. (Christou & Koenderink, 1997 ) The above left bias was established in 2003 in which psychological studies found that human observers assume that the light source is coming from the above left. (Mamassian et al, 2003. ,) It has also been established that the observers prefer the stimuli lit from the left rather than the right (Geradin, Montalembert & Mamassian, 2007) Another established bias in establishing shape from shading is a convex bias where humans were more likely to precieve 3D shapes in different lighting settings as convex rather than concave. It was also revealed that convex surfaces produce a greater perceived depth than concave surfaces as a part of this bias. (Liu, & Todd, 2004) This bias has been shown to be strong when the stimuli is blurred. (Geradin, Montalembert & Mamassian, 2007)

Current
More modern research on shape from shading looks to develop new technologies such as |kwd-294717207|cid|16055868139|aid|579112695766|gid|133367713672|pos||src|g_|dvc|c|reg|1007409|rin||fid|&utm_term=facial%20recogn face recognition AI (Zhao & Chellappa, 2000) Shape from shading  Shape from shading research is being used to improve technology in differentiating between 2D and 3D face recognition in different light settings.

This technology can be applied in society to prevent crimes such as spoofing. (Martino et al, 2020) which uses facial recognition technology. Security aspects of biometric systems are essential for a successful authentication mechanism due to the large possibility that an impostor user has for attacking it. A visual attack would be the most simple way to elude such systems. This could be performed by presenting a synthetic sample, such as a photograph, a video or even a 3D mask of a registered user to the sensor to evade the security measures and gain access. (pinto et al, 2020) Anyone who has access to a photo of a person with valid access to a system with relies of face presentation security could use it to break into the system. The best software for detecting face presentation attacks found a attack presentation classification error rate of 5% under environmental, attack types and camera device variations. This means that 5/100 attacks were successful which may in future lead to the facial recognition process no ;longer feasible for the future, especially if a system has hundreds of users. This would leave companies vulnerable and any sensitive information could be exposed.(Martino et al, 2020)