US 11,308,582 B2 · Blended neural network for super-resolution image processing

Abstract

Embodiments relate to a super-resolution engine that converts a lower resolution input image into a higher resolution output image. The super-resolution engine includes a directional scaler, an enhancement processor, a feature detection processor, a blending logic circuit, and a neural network. The directional scaler generates directionally scaled image data by upscaling the input image. The enhancement processor generates enhanced image data by applying an example-based enhancement, a peaking filter, or some other type of non-neural network image processing scheme to the directionally scaled image data. The feature detection processor determines features indicating properties of portions of the directionally scaled image data. The neural network generates residual values defining differences between a target result of the super-resolution enhancement and the directionally scaled image data. The blending logic circuit blends the enhanced image data with the residual values according to the features.

Claims20

Dependent claims

An image signal processing device, comprising:

a feature detection processor configured to analyze an input image to identify image data of a portion of the input image as including a feature that a neural network model is trained to handle for resolution enhancement, the portion of the input image having a first resolution;
a neural network circuit that implements the neural network model, the neural network circuit configured to generate enhanced image data from an upscaled version of the portion of the input image, the upscaled version having a second resolution higher than the first resolution of the portion of the input image;
and
a blending logic circuit coupled to the feature detection processor and configured to generate an output image by applying weights to values of the enhanced image data in the output image based on the feature detected by the feature detection processor, the output image having a third resolution that is higher than the first resolution at least with respect to the portion of the input image.

1 more element

Dependencies are inferred from the claim text.

Blended neural network for super-resolution image processing

Life of this patent

Blended neural network for super-resolution image processing

Life of this patent

Abstract

Claims20