发明名称 Image processing apparatus for generating an image for three-dimensional display
摘要 A detecting unit detects an object in an input image. A depth map generating unit selects a depth template corresponding to a type of the object and places a selected depth template on a depth map in accordance with a position of the object to generate the depth map having a depth value for each pixel. A correcting unit calculates a weight of at least one interested pixel and a weight of a peripheral pixel based on a relationship between pixel values to the interested pixel and the peripheral pixel and corrects the depth value of the interested pixel based on a weighted sum of the respective depth values corresponding to the interested pixel and the peripheral pixel. An image generating unit generates parallax images based on the corrected depth map and the input image.
申请公布号 US9053575(B2) 申请公布日期 2015.06.09
申请号 US201113234448 申请日期 2011.09.16
申请人 Kabushiki Kaisha Toshiba 发明人 Shimoyama Kenichi;Mishima Nao;Mita Takeshi;Ida Takashi
分类号 H04N13/02;G06T15/20 主分类号 H04N13/02
代理机构 Nixon & Vanderhye, P.C. 代理人 Nixon & Vanderhye, P.C.
主权项 1. An image processing apparatus for a moving image having multiple frames, the image processing apparatus comprising: a processing system configured to: detect an object in an input two-dimensional image, corresponding to a current frame, including pixel values of each pixel and a size and an orientation of the object; select, from at least one depth template defining a depth value of a pixel, a depth template corresponding to a type of the object, modify the selected depth template based on the size and orientation of the object, and place the modified depth template on a depth map in accordance with a position of the object to generate the depth map having a depth value for each pixel of the input image, wherein the processing system modifies the selected depth template by changing the depth value of the selected depth template based on a difference between an orientation of the selected depth template and the detected orientation of the object if the detected orientation of the object is different from the orientation of the selected depth template; calculate a weight of at least one interested pixel and a weight of a peripheral pixel in the depth map based on a relationship between respective pixel values corresponding to the interested pixel and the peripheral pixel and provide a corrected depth map by correcting the depth value of the interested pixel in the depth map based on a weighted sum of respective depth values of the interested pixel and the peripheral pixel; combine the corrected depth map and another depth map to provide a combined depth map, the other depth map being a depth map for an input two-dimensional image corresponding to a frame t frames before the current frame; and generate a plurality of parallax images based on the combined depth map and the input two-dimensional image, wherein the processing system is further configured to determine a median value among pixel values corresponding to the interested pixel and peripheral pixels, assign a weight of 0 to a pixel associated with a pixel value that does not match a median value out of the peripheral pixels and the interested pixel, andcalculate an average of the depth value of a pixel associated with a pixel value that matches the median value out of the peripheral pixels and the interested pixel and correct the depth value of the interested pixel to the average.
地址 Tokyo JP