Image region segmentation method and system using self-spatial adaptive normalization

An image region segmentation method comprising:

• encoding an image for segmenting a region by using a plurality of encoding blocks;
• and
• decoding the encoded image and generating a region-segmented image by using a plurality of decoding blocks,
• wherein each of the encoding blocks is configured to process an inputted image into a convolution layer, to perform spatial adaptive normalization, and then to reduce the image and to deliver the image to the next encoding block.

An image region segmentation system comprising:

• an encoder configured to encode an image for segmenting a region by using a plurality of encoding blocks;
• and
• a decoder configured to decode the image encoded by the encoder and to generate a region-segmented image by using a plurality of decoding blocks,
• wherein each of the decoding blocks comprises:
• an up-sampling block configured to increase an output of the encoder;
• a convolution block configured to receive an output of the up-sampling block and an output of a normalization block included in the encoding block, and to process the output into a convolution layer;
• and
• a normalization block configured to perform spatial adaptive normalization with an output of the convolution block and to deliver the output to the next decoding block.