Hierarchical Interlinked Multi-scale Convolutional Network for Image Parsing

A method for parsing an image into components, the method comprising:

• generating N levels of image patches from the image, N>2, wherein the image patches or different levels n have different resolutions R(n) and the image patches for a level n are generated from the image resampled to resolution R(n);
• generating M levels of interlinked image patches from the N levels of image patches, 2<M≦N, wherein the interlinked image patches for different levels m have different resolutions R(m) and the interlinked image patches for a level m are generated from an input group m of image patches comprising (i) image patches from level n with R(n)=R(m), and (ii) image patches from one or more levels n with R(n)≠R(m) where such image patches have been resampled to resolution R(m);
• and
• locating the components by combining the M levels of interlinked image patches.

A non-transitory computer readable medium configured to store program code, the program code comprised of instructions for parsing an image into components, the instructions when executed by a processor cause the processor to:

• generate N levels of image patches from the image, N>2, wherein the image patches or different levels n have different resolutions R(n) and the image patches for a level n are generated from the image resampled to resolution R(n);
• generate M levels of interlinked image patches from the N levels of image patches, 2<M≦N, wherein the interlinked image patches for different levels m have different resolutions R(m) and the interlinked image patches for a level m are generated from an input group m of image patches comprising (i) image patches from level n with R(n)=R(m), and (ii) image patches from one or more levels n with R(n)≠R(m) where such image patches have been resampled to resolution R(m);
• and
• locate the components by combining the M levels of interlinked image patches.