In a section of the present disclosure, a system for generating a 3D image from a 2D image and a depth map of the 2D image is disclosed. A server receives the 2D image from a client device. A minimum depth value is set for pixels at bottom of the 2D image. An object detection algorithm identifies presence of an object in the 2D image. A maximum depth value is set at a position of the object, if it is detected; otherwise, it is set at top of the 2D image. A continuous depth function varying from the minimum depth value to the maximum depth value is used to represent the depth map. A stereoscopic image is then generated utilizing the depth map of the 2D image. Thereafter, a parallax disparity in the stereoscopic image is generated to produce two views in the stereoscopic image. Finally, the two views in the stereoscopic image are used to generate the 3D image.
In another section of the present disclosure, a system combines at least two input images (at least one is a 3D image or has a depth map) into a new combined 3D image. A server receives the input images (for example, a first RGBD image, a second RGBD image and an RGB image) from a client device. A first depth map is extracted from the first RGBD image and a second depth map is extracted from the second RGBD image. Thereafter, a depth map is also generated for the RGB image, referred to as a third depth map. Finally, a combined depth map is generated from the first depth map, the second depth map and the third depth map. In a next step, weight masks are defined for the input images. Each of the weight masks represents a set of coefficients. The set of coefficients are calculated by performing a histogram alignment and by applying a mismatch bias based on the histogram alignment. The weight masks are applied to pixels in each of the input images by multiplying the set of coefficients with values of the pixels at each pixel location. Then the input images with the applied weight masks are combined to produce a combined RGB image. Thereupon, the server generates the new combined 3D image from the combined RGB image and the combined depth map.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Anonymous, Anonymous, "Multi-Input 3D Photos", Technical Disclosure Commons, (August 06, 2019)