abstract |
The invention discloses a multi-instance image segmentation method based on coding auxiliary information, which belongs to the technical field of image coding and instance segmentation. Aiming at the deficiencies caused by the existing multi-instance image segmentation methods that only use the original image to perform instance segmentation, the present invention provides a multi-instance image segmentation method based on coding auxiliary information. The present invention obtains luminance and color-difference macroblocks of different sizes through an image decoding algorithm for an input image, and extracts intra-frame prediction direction information, so that the obtained coding unit scale spectrum and intra-frame prediction direction spectrum are used as coding auxiliary information, which can make full use of information about the image. The invention uses the long short-term memory network for fields other than text classification and natural language processing, and uses the long short-term memory network to fuse the scale spectrum, the direction spectrum and the original image together, thereby improving the accuracy of multi-instance image segmentation. |