Depth Information Estimation of Image Texture Direction and Spatial Distribution Features

The acquisition of image depth information is a key component of computer vision and graphics image recovery. However, the disadvantages of the expensive dual-image depth information acquisition apparatus and the complexity of image matching are shortcomings. Therefore,this thesis focuses on the deep research of single image depth information.Firstly, this thesis obtains the depth information from the texture structure of a single image. It not only considers the feature of the direction change of the texture, but also takes into account the spatial distribution of the texture, and can fully and completely describe different types of images. Secondly, it divides the pixels in the image. The feature information is extracted from image blocks in multi-scale space. Then the pixels in the L-shape neighborhood are used as the primitives to describe their characteristics and distribution rules. The model is modeled by MRF and Gibbs. Finally, the MRF model and texture are used. The features are combined to estimate the depth information of the image. Experiments show that this method is cheap, efficient, and simple and easy to implement.