Method for obtaining image depth information

Abstract

The invention discloses a method for obtaining image depth information. The method for obtaining the image depth information includes the following steps that firstly, a single-width primitive image to be processed undergoes Gaussian Blur processing to obtain a blurred image; secondly, a grain rim of the primitive image to be processed is detected, the primitive image is divided into a zone with the large grain gradient which is defined as a D zone and a zone with the large grain gradient which is defined as an F zone; thirdly, in terms of pixel points in the D zone, scale factors of all the pixel points are obtained by calculation with a blurred estimation method; fourthly, in terms of all the pixel points in the F zone, Kalman filtering is carried out and scale factors of all the pixel points are estimated; fifthly, according to focusing information of the primitive image, the scale factors of all the pixel points are converted into relative depth values of all the pixel points. According to the method for obtaining the image depth information, the idea of Kalman filtering is introduced, noise generated when depth values are obtained with a regular defocusing method can be hindered, the precision of the final depth image is improved and extra information supplementing is not needed.

Description

【Technical field】 [0001] The invention relates to the field of computer vision, in particular to a method for obtaining depth information from images. 【Background technique】 [0002] Computer vision technology is to obtain input by simulating the human eye through imaging systems such as cameras, and the computer simulates the human brain for processing and interpretation, so that the computer can observe and understand the world through vision like a human. In computer vision, in order to make computer vision work like human vision, a basic problem is how to obtain the structure and attribute information of the three-dimensional world from the two-dimensional images captured by the camera? That is, from the two-dimensional images captured, Extract depth information and obtain a depth map. For this reason, many methods have been proposed, and according to different sources of depth information, they can be divided into two categories: active vision and passive vision. [0...

AI Technical Summary

Problems solved by technology

However, due to the interaction between the depth cues of the defocus method and the content and noise of the image, the conventional single-image defocus method to obtain the depth map pays less attention to the noise reduction, so the obtained depth map The accuracy is often limited

Images