National Cheng-Kung University

Gwo Giun (Chris) Lee

He-Yuan Lin

Chun-Fu Chen

Ming-Jiun Wang

Ping-Keng Jao

A method for depth map generation is disclosed, capable of generating a depth map corresponding an image signal, for the application of a 2D to 3D image transformation system. In the depth map generated by the disclosed method, each of the plural image regions of the image signal is assigned with a depth value. Besides, by means of comparing the depth map with another depth map of the earlier time point, the disclosed method can generate a modulated depth map, for assigning a depth value to each of the plural image regions of the image signal more precisely. Thus, the transformation performance and efficiency of the 2D to 3D image transformation system are hereby improved.

Method for depth map generation

The present invention relates to a method for merging regions in the image/video, capable of merging plural of image regions into an image merging region. In the disclosed method, these image regions are first sequenced basing on their compactness value. Then, one of these image regions is designated as a reference image region, and a merging test process is executed by merging the reference image region with one of the nearby image regions thereof in sequence, for forming a temporal image merging region. Later, the compactness value of the temporal image merging region is compared with the compactness value of the two consisting image regions thereof, respectively. When the compactness value of the temporal image merging region is larger than either one of the compactness value of the two consisting image regions thereof, the temporal image merging region is designated as an image merging region.

Method for merging the regions in the image/video

A method for color feature extraction extracts a color feature vector representative of the color of each image pixel contained in an image signal. The method comprises: receiving the image signal; mapping the image signal to a color space model, where the color of each of the plural image pixels is represented by a first parameter, a second parameter, and a third parameter; obtaining an adjusted second parameter; clustering the plural image pixels into plural color regions or plural fuzzy regions of a color plane of the color space model; and designating the color feature vector to each of the plural image pixels based on the clustering result.

Method for color feature extraction

A method for image/video segmentation, capable of segmenting an image signal for obtaining plural texture color feature regions, by utilizing both of the advantages carried by the texture feature and the color feature is disclosed. The method comprises the following steps: (A) receiving an image signal including plural image pixels; (B) executing a Gabor filtering process and a value operation process on each of the plural image pixels; (C) designating each of the plural image pixels a corresponding texture feature vector basing on the result of the value operation process; (D) executing a segmentation process on the image signal basing on the texture feature vector of each of the plural image pixels, for obtaining plural texture feature regions; and (E) executing a re-segmentation process on plural color feature regions basing on the distribution of the plural texture feature regions, for obtaining plural texture color feature regions.

Method for image/video segmentation using texture feature

A method for 3D video content generation is disclosed, capable of transforming a 2D image into a 3D video through proper operation process. The method comprises the following steps of: (A) receiving a 2D image and generating a ROI distribution map from the 2D image; (B) executing a color feature capture process, for forming a plural of color feature regions; (C) executing an image segmentation process basing on the texture feature of the plural of color feature regions, for forming an image region distribution map; (D) executing a depth map generation process, for generating a depth map basing on the ROI distribution map and the image region distribution map; (E) executing a 3D image generation process, for forming the 3D image basing on the image region distribution map and the depth map; and (F) chaining a plurality of the 3D images to form the 3D video basing on a frame rate.

Growing community of inventors

Tainan, Taiwan

Gwo Giun (Chris) Lee