IMAGE SEGMENTATION METHOD AND IMAGE PROCESSING DEVICE

WIPO Patent Application WO/2020/177651

A1

The present application discloses an image segmentation method in the field of artificial intelligence. The method comprises: obtaining an input image and a processing requirement; performing multi-level feature extraction on the input image so as to obtain a plurality of feature maps; downsampling the plurality of feature maps so as to obtain multiple feature maps having reference resolutions lower than the resolution of the input image; fusing the multiple feature maps having the reference resolutions, so as to obtain at least one feature map group; upsampling the feature map group using a transformation matrix W so as to obtain a target feature map group; and performing target processing on the target feature map group according to the processing requirement so as to obtain a target image. In the present application, multiple lower-resolution feature maps are fused to obtain a feature map group, and the feature map group is then upsampled by means of a transformation matrix W so as to obtain a target feature map group, thereby effectively reducing memory footprints and calculations, and improving the accuracy of image segmentation.

TIAN ZHI (AU)
HE TONG (AU)
SHEN CHUNHUA (AU)
YAN YOULIANG (CN)
XU SONGCEN (CN)
ZHOU YIREN (CN)
WU XIAOFEI (CN)
LIU JIANZHUANG (CN)

PCT/CN2020/077366

September 10, 2020

March 01, 2020

Click for automatic bibliography generation  

HUAWEI TECH CO LTD (CN)

G06K9/34

CN110033003A	2019-07-19
CN104794504A	2015-07-22
CN109034162A	2018-12-18
CN108876793A	2018-11-23

SCIHEAD IP LAW FIRM (CN)

View/Download PDF      PDF Help

Previous Patent: BREAKER DEVICE AND INVERTER SYSTEM

Next Patent: METHOD AND SYSTEM FOR PRODUCING LIGHT OLEFIN FROM POOR OIL