The present invention relates to a real-time multi-view image generating method based on kinect which includes the steps of: obtaining a depth image using an infrared pattern of the kinect; generating a virtual view image by correcting a blurring phenomenon generated around a boundary; and filling a hole of an occlusion region of the virtual view image based on reliability and a slope based on a block. The present invention includes the steps of: (a) obtaining the depth image and a color image; (b) filling the hole using a joint bilateral filter in the depth image; (c) searching and correcting the blurring pixel of the depth image; (d) generating the virtual view image using 3D warping; and (e) filling the hole of the occlusion region in the virtual view image. According to the multi-view image generating method, a multi-view image with high quality is generated by reducing the blurring phenomenon of the screen, preventing wrong color information from being used in a hole filling process, separating a background region from an object region, and maintaining the continuity of the image through the slope information of the image in comparison with an existing method.
展开▼