3D TV Image Generation Method

3D TV has received more and more attention, and many businesses have also launched some related products. So how exactly did this happen? I will introduce some of the knowledge points that I have touched to you here.

IResearch status at Home and Abroad

The best way to use a Three-Dimensional TV is to use two cameras to take the current scene at the same time according to a certain angle, and then combine the images taken at the same time, this will create an image that looks a little out of stock and blurred to the naked eye. We call the two images left and right respectively. If you have a special pair of glasses or TV screen, you can see the Three-Dimensional TV image. However, in reality, two cameras are usually used and taken from a certain angle, which has many difficulties. Therefore, in reality, there is only one image, and we can find a way to generate another corresponding image.

There are few documents in China about this and it is also a very inaccurate solution. For example, assume that the vast majority of scene images are in the upper half of the image as the foreground, and the lower half as the near scene. Then, according to the top to bottom, the image is gradually shifted to the left or right, however, not all scenarios are in this situation. In addition, the method of adding the offset factor to the Image Based on the psychological and physiological factors is also unsatisfactory.

Method for generating Binary Images

1. Image Segmentation

Currently, one image is used to generate another image. A more accurate method is to use the image segmentation method to retrieve all the regions in the image and then based on the image depth, the corresponding displacement of each part makes the image effect reasonable.

There are also many methods for image segmentation. The author uses the principles based on rough set theory. The specific implementation process will not be detailed here. If you are interested, leave a message to discuss it. Directly:

Fig1 original image Fig2 added depth after segmentation Image

2. depth chart

The split image cannot be used directly for the left or right image. We also need to assign a depth value to each segment of the Image Based on the image type. In general, we can determine the depth of an image by directly detecting the image. For example, if a horizontal line (the middle path) can be detected in the first row of a graph, it is determined that it belongs to the preceding foreground and the bottom is the near scene type. This is determined by the straight line direction. If it cannot be determined, the default image belongs to the Top View and the bottom view type.

3. Split block displacement

Displacement Formula:

XcIs the horizontal coordinate of the intermediate view. ZIs depth value of current pixel, FIs camera focal length and T xIs eye distance.4 Three-Dimensional Image Generation Survey found that about 70% of people belong to the right eye sensitivity, 20% belong to the left eye sensitivity, 10% of people around the eyes are similar. Therefore, we can use the original image as the right image and generate the left image through Segmentation and depth.

After the three figures are combined, the corresponding three-dimensional graph is generated:

Figure 5 stereoscopy

Other examples: