FEATURE-PRESERVING NOISE REMOVAL

摘要

A noise removal method includes receiving in a de-noising system multiple noisy copies of a target image; providing the noisy copies as input data to the de-noising system in a training phase, and determine a noise distribution for the plurality of noisy copies. The method further includes, for n stages of the de-noising system, transforming the input data into an output; reconstructing the output; and providing the reconstructed output as input data. The method may further include receiving a new image and generating a de-noised image based on the determined noise distribution.

主权项

1. A noise removal system comprising a processor, the processor configured to perform a first stage of reconstruction including a first portion, the first portion of the first stage of reconstruction comprising to:
receive a plurality of copies of a data space; for each copy of the plurality of copies, create an input vector representing the copy, the input vector created by applying a window to successive locations of the copy, and for each of the successive locations, transforming data within the window into a single data point of the input vector; provide a set of input vectors, including an input vector for each copy of the plurality of copies, to a first stage neural network; receive from the first stage neural network an output vector representing the set of input vectors; and generate from the output vector a reconstructed image.