This publication describes systems and techniques to enhance an image quality of a photograph taken by a portable device using a collection of images to find a replacement object. Although the intrinsic image quality of photographs taken by portable devices, such as a smartphone, continues to improve, many pictures still lack detail or have blurriness. Consequently, a photograph, or “original image,” can include an object having a poor quality, such as blurriness or an inaccurate reproduction of a texture. To enhance the original image, a machine-learned model detects in the original image the object having poor quality. A collection of images is searched to find a reference image having another version of the object with a superior quality. Metadata for the images can be used to find a matching reference image. Metadata can include positioning, distance to subject, tilt angle, and so forth. An object enhancement module replaces at least a portion of the poor-quality object in the original image with the superior quality object from the reference image to produce an enhanced image. The image collection with reference images can be stored in the cloud or locally on the portable device. In these manners, the image quality of a photograph taken by a portable device can be enhanced.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Chen, Yen-Cheng, "Enhancing Image Quality of Photographs Taken by Portable Devices by Matching Images to High Quality Reference Images Using Machine Learning and Camera Orientation and Other Image Metadata", Technical Disclosure Commons, (November 25, 2019)