A method of image depth estimation based on light field

Abstract

The invention relates to an optical field based image depth estimation method. The method mainly comprises three important parts of an optical field original data view point extraction method, a block matching based depth estimation algorithm and a remarkable feature constraint based depth optimization algorithm. The proposed optical field original data view point extraction method is used for performing view point separation on optical field data not subjected to demosaicing. The adopted block matching based depth estimation algorithm is used for performing similarity measurement on view points located in the same row or column as a central view point and corresponding blocks that record same light ray colors. For optimizing a depth estimation result, the invention proposes the remarkable feature constraint based depth optimization algorithm for extracting and matching remarkable feature points and taking a parallax error of the remarkable feature points as a strong constraint condition. The method avoids view point confusion caused in an interpolation process and improves depth estimation accuracy.

Description

technical field [0001] The invention relates to an image depth extraction method, in particular to an image depth estimation method based on a light field. Background technique [0002] Image depth estimation is a fundamental topic in the field of computer vision. Depth refers to the distance from a point in the scene to the camera plane. If the depth of the scene can be accurately recovered from the real image, it will be beneficial to the realization of many computer vision applications. In recent years, the emergence of light field cameras has proposed new solutions for image depth estimation. Compared with traditional cameras, light field cameras place a microlens array in front of the sensor, which can record the position and angle of any light that reaches the imaging surface during one exposure, comprehensively describe the four-dimensional light field, and realize depth estimation, Various applications such as scene refocusing and viewpoint change. [0003] At pr...

AI Technical Summary

Problems solved by technology

For example, mainstream light field image depth estimation methods focus on depth estimation of light field data from separated viewpoints. Demosaiced light field data are used for viewpoint separation. However, the confusion between viewpoints introduced during the demosaicing process cannot It is removed in subsequent processing, which greatly restricts the accuracy of depth estimation

In addition, some algorithms use empirically calculated confidence levels for depth estimates, adding artifacts to the optimization process

Images