System and method for automatic digital image orientation detection
AssigneeFuji Xerox Co., Ltd.
InventorsDong Wang, Matthew L. Cooper
Life of this patent
Estimating the term…
Abstract
Described is a system for automatic digital photo orientation detection. We leverage online public photos with great content variation to extract effective features with layout information. Classification proceeds using an approximate nearest neighbors approach which scales well to massive training sets, hardly compromising efficiency. We have tested the method successfully on the largest data set to date of nearly 30,000 Flickr photos as well as both difficult and typical consumer usage scenarios. Though limited data are available for comparison across different systems, the proposed system significantly outperforms a state of the art system on a common data set.
Claims22
A method for determining an optimal image orientation, the method comprising:
- a. Obtaining an input image;
- b. Extracting image features for the input image;
- c. Generating rotated image features corresponding to alternate orientation of the image;
- d. Analyzing the extracted and generated features using a plurality search trees constructed from large sets of images of known orientation, the plurality of search trees comprising candidate nearest neighbors;
- e. Combining orientations of the candidate nearest neighbors using a weighted vote;
- and
- f. Determining the optimal image orientation of the input image using a result of e.
A computer readable medium embodying a set of instructions, the set of instructions, when executed by one or more processors, is operable to cause the one or more processors to perform a method for determining an optimal image orientation, the method comprising:
- a. Obtaining an input image;
- b. Extracting image features for the input image;
- c. Generating rotated image features corresponding to alternate orientation of the image;
- d. Analyzing the extracted and generated features using a plurality search trees constructed from large sets of images of known orientation, the plurality of search trees comprising candidate nearest neighbors;
- e. Combining orientations of the candidate nearest neighbors using a weighted vote;
- and
- f. Determining the optimal image orientation of the input image using a result of e.
A system for determining an optimal image orientation, the system comprising:
- a. An input module operable to obtain an input image;
- b. An image feature extracting module operable to extract image features for the input image;
- c. A rotated image features generation module operable to generate rotated image features corresponding to alternate orientation of the image;
- d. A search tree module operable to analyze the extracted and generated features using a plurality search trees constructed from large sets of images of known orientation, the plurality of search trees comprising candidate nearest neighbors;
- and
- e. A combining module operable to combine orientations of the candidate nearest neighbors using a weighted vote;
- wherein the optimal image orientation of the input image is determined based on a result of the combining.
Dependencies are inferred from the claim text.