Method and system for three-dimensional model reconstruction

A method of generating a three-dimensional model of an object by a light field camera configured to capture a plurality of light field images at a plurality of viewpoints, the method comprising:

• capturing a first light field image at a first viewpoint, wherein the first light field image includes a first plurality of subaperture images;
• capturing a second light field image at a second viewpoint, wherein the second light field image includes a second plurality of subaperture images;
• estimating a rotation and a translation of a light field from the first viewpoint to the second viewpoint;
• obtaining a disparity map from each of the plurality of light field images;
• computing a three-dimensional point cloud by optimizing the rotation and translation of the light field and the disparity map;
• identifying a plurality of first features on the first light field image and a plurality of second features on the second light field image;
• matching the first features across the first plurality of subaperture images to obtain first matched features, and matching the second features across the second plurality of subaperture images to obtain second matched features;
• matching the first matched features to the second matched features between the first light field image and the second light field image to obtain matched features between the light field at the first viewpoint and the second viewpoint;
• and
• estimating a rotation and a translation of the light field based on the matched features between the light field at the first viewpoint and the second viewpoint.

A system for generating a three-dimensional model of an object, comprising:

• a light field camera configured to capture a plurality of light field images at a plurality of viewpoints;
• and
• a data processing unit for generating the three-dimensional model,
• wherein a first light field image which includes a first plurality of subaperture images, is captured at a first viewpoint, a second light field image which includes a second plurality of subaperture images, is captured at a second viewpoint, and the data processing unit is configured to estimate a rotation and a translation of a light field from the first viewpoint to the second viewpoint, obtain a disparity map from each of the plurality of light field images, and compute a three-dimensional point cloud by optimizing the rotation and translation of the light field and the disparity map,
• and wherein the plurality of viewpoints cover all viewpoints of the object.