The Method Of Image Depth Estimation Based On Texture Feature Probability Model

The3D film promote the research of image depth information extraction method due to the attraction of depth information. The development of the research about image depth information affects many industries such as the TV and the movie industry. As the intelligent ideas become more and more popular nowadays, it will spread rapidly in the near future for the intelligent equipment such as smart home furnishing, intelligent traffic, and intelligent vehicle etc, and the image depth information extraction technology will directly affect the intelligent level of these applications. In addition, this technology can be also applied to image restoration and reconstruction about the national cultural relics and ancient buildings, so it can protect the cultural heritage.By analyzing and comparing different depth cues, we find each depth cue has its own application scope, but texture cue have the advantage of strong applicability and less restriction. Therefore, we choose texture cue as the main feature of the image, we use Laws’filters to extract image texture gradients, texture variations and colors respectively. After comparing different scales’location, size, direction, we can estimate this texture region’s relative position and get its deep information.Based on the study of texture gradient characteristics and optimization algorithm, We proposed a single image depth estimation method based on markov random field (MRF) and multi-scale texture features, which uses Laws filers to the two dimensional image to calculate the feature of texture gradients, texture variations and colors, and it attributes the depth estimation problem to pattern recognition problem, i.e., it divides the depth of the two-dimensional image into many classes of depths, and calculate the probability relation between texture clues and scene depth according to the texture characteristics at different scales, then, it builds MRF probabilistic model to get the initial depth image. In order to improve the quality of the depth image, we recalculate the image depth information through the iterating algorithm on the basis of the initial depth we obtained. The experimental results show that it is effective by adopting different iterative criterion.