Jia Xu

Title: Jia Xu: Inventor in Image Processing

Introduction

Jia Xu is an inventor based in Santa Clara, California, whose patent work focuses on image processing and computer vision. She is associated with Intel Corporation and has 2 patents.

Latest Patents

- Optical flow estimation using 4-dimensional cost volume processing. This invention describes estimating motion between images by extracting feature vectors, building a four-dimensional cost volume from distances between selected features, and applying flow-semi-global matching to estimate optical flow vectors.

- Approximating image processing functions using convolutional neural networks. This invention uses a trained convolutional neural network to approximate a targeted image-processing operator. The network learns bias and convolutional kernel parameters from training image pairs, then applies a sequence of nonlinear operations to produce an output image.

Career Highlights

Xu’s listed patent contributions address computational methods for image analysis and processing. Her work connects optical-flow estimation with neural-network-based approximations of image-processing functions.

Collaborations

Her listed coworkers include Vladlen Koltun and Qifeng Chen.

Conclusion

Jia Xu’s patents reflect work on computer-vision techniques that use cost-volume matching and convolutional neural networks to process images.