Image quality enhancing
AssigneeBeijing Baidu Netcom Science Technology Co., Ltd.
InventorsJiayin Cai, Huaifei Xing
Life of this patent
Estimating the term…
Abstract
The present disclosure provides a method and an apparatus for enhancing image quality, a device, and a medium, relates to the field of artificial intelligence and specifically to computer vision and deep learning technologies, and can be applied to an image processing scenario. The method includes: determining an ROI and an RONI in an image to be processed; inputting the ROI to an ROI image quality enhancement model, to obtain first image data output from the ROI image quality enhancement model; inputting the RONI to an RONI image quality enhancement model, to obtain second image data output from the RONI image quality enhancement model; and blending the first image data and the second image data.
Claims14
A method for enhancing image quality, comprising:
- determining a region-of-interest (ROI) and a region-of-non-interest (RONI) in an image to be processed, wherein the ROI is a foreground object determined by performing one or more of object detection or image semantic segmentation on the image to be processed, and the RONI is a region of the image to be processed excluding the ROI;
- selecting, from a model library, an ROI image quality enhancement model and an RONI image quality enhancement model at least based on an expected speed for processing the image to be processed, wherein the ROI image quality enhancement model uses a deep-learning method, and the RONI image quality enhancement model uses a non-deep-learning method, wherein the non-deep-learning method comprises at least one of:
- smoothing filtering, sharpening, histogram equalization, image morphology, and wavelet transform, wherein complexity of the ROI image quality enhancement model selected for processing the image to be processed is greater than complexity of the RONI image quality enhancement model selected for processing the image to be processed;
- inputting the ROI to the ROI image quality enhancement model, to obtain first image data output from the ROI image quality enhancement model;
- inputting the RONI to the RONI image quality enhancement model, to obtain second image data output from the RONI image quality enhancement model;
- and
- performing a Poisson blending on the first image data and the second image data.
An electronic device, comprising:
- one or more processors;
- a memory storing one or more programs configured to be executed by the one or more processors, the one or more programs comprising instructions for:
- determining a region-of-interest (ROI) and a region-of-non-interest (RONI) in an image to be processed, wherein the ROI is a foreground object determined by performing one or more of object detection or image semantic segmentation on the image to be processed, and the RONI is a region of the image to be processed excluding the ROI;
- selecting, from a model library, an ROI image quality enhancement model and an RONI image quality enhancement model at least based on an expected speed for processing the image to be processed, wherein the ROI image quality enhancement model uses a deep-learning method, and the RONI image quality enhancement model uses a non-deep-learning method, wherein the non-deep-learning method comprises at least one of:
- smoothing filtering, sharpening, histogram equalization, image morphology, and wavelet transform, wherein complexity of the ROI image quality enhancement model selected for processing the image to be processed is greater than complexity of the RONI image quality enhancement model selected for processing the image to be processed;
- inputting the ROI to the ROI image quality enhancement model, to obtain first image data output from the ROI image quality enhancement model;
- inputting the RONI to the RONI image quality enhancement model, to obtain second image data output from the RONI image quality enhancement model;
- and
- performing a Poisson blending on the first image data and the second image data.
Dependencies are inferred from the claim text.