Methods and systems for 2D to 3D conversion from a portrait image

摘要

A method for converting a 2D image into a 3D image includes receiving the 2D image; determining whether the received 2D image is a portrait, wherein the portrait can be a face portrait or a non-face portrait; if the received 2D image is determined to be a portrait, creating a disparity between a left eye image and a right eye image based on a local gradient and a spatial location; generating the 3D image based on the created disparity; and outputting the generated 3D image.

主权项

1. A computer-implemented method for converting a 2D image into a 3D image, the method comprising:
receiving the 2D image; determining whether the received 2D image is a close-up image, wherein the close-up image is a face portrait or a non-face close-up image, wherein the close-up image includes a sharp foreground and a blurred background; if the received 2D image is determined to be a close-up image, creating a disparity between a left eye image and a right eye image based on a local gradient and a spatial location; generating the 3D image based on the created disparity; outputting the generated 3D image; determining whether the received 2D image is a non-face close-up image, the determining comprises:
detecting that the received 2D image is not a face portrait;dividing the received 2D image into a plurality of cells;segmenting foreground cells containing a foreground object from background cells in the plurality of cells, the segmenting comprises:
choosing a first foreground cell based on sharpness values of pixels in the first foreground cell and sharpness values of pixels in the plurality of cells excluding the first foreground cell, anditeratively selecting a next foreground cell adjacent to selected foreground cells based on sharpness values of pixels in a candidate foreground region, sharpness values of pixels in a candidate background region, and sharpness values of pixels in the plurality of cells;extracting features from the segmented foreground and background cells; andclassifying the received 2D image as a non-face close-up image based on the extracted features.