The Research Of Self-shadow Algorithm Based On Gray Image

Shape from shading(SFS for short) is an important method for three-dimensionalreconstruction techniques (Shape from X for short, X may be movement, texture,outline, shadow, etc.). It is able to restore the three-dimensional shape of the objectfrom the gray degree information by a single image. The technology is relatively easyto implement, low cost, it not require the high the hardware device and the externalconditions. And it can be used in some special occasions where other techniques arenot applicable. With the continuous development of computer vision and virtualreality,3D reconstruction of images is becoming an important study day by day, andplays more and more important status and role in the various fields of production andlife in society, especially in industrial, medical, national defense, archeology, etc. Itscorresponding problem also has an extremely widely value of applications.It is very common in real world that the shadow will appear as long as an object isin front of the light source. Generally, shadows are composed of two parts: selfshadows and cast shadows. The cast shadow is projected in the scene object underillumination of the shadow region; it is independent of the objects themselves.However, the self shadow is the part of the object, which is not illuminated by thelight source. From self-shadows of its own particularity, the former three-dimensionalmethod based on gray image treat to the self-shadowing problems always been to takeevasive strategy which will result in its three-dimensional reconstruction shapedistortion, thus affecting the effect and accuracy of the three-dimensional methodbased on gray image.To solve the problem of self-shadows, this paper mainly study the algorithm ofself-shadows based on the edge information and the analysis method of regionalgradient. In this paper, we improve and optimize the existing algorithms. Under thepremise of light source direction has been estimated, detect shadow area and objectwith self-shadows area by edge information, and use the light source projectedcoordinate system to analyze self-shadows, then use the analysis method of regionalgradient analyze self-shadows area for further processing. Finally, we use realgray-scale images to verify this algorithm in this paper.The results of the experiments verify that the algorithm in this paper is effective.This algorithm can well deal with the problem of self-shadows and can be applied tothe objects with self-shadow and more complicated shape.