Abstract: An image encoding/decoding method, device and recording medium based sed on multiple compression levels disclosure may include extracting a region of interest for machine vision from an input image, determining a compression level of the region of interest, and encoding the compression level of the region of interest.

Abstract: An object-based image encoding/decoding method and device of the present disclosure may include an image partition step which partitions an image to obtain a first object region, a region scaling step which scales the first object region based on a scaling factor of the first object region to obtain a second objection region, a region merging step which merges the second object region with at least one of an object region different from the second object region or a non-object region to obtain a merged image, and an image reconstruction step which reconstructs the merged image.

Abstract: An image encoding method includes detecting at least one attention region from an original image, acquiring a reconstructed picture for the original image, generating residual data between a first attention region in the original image and a second attention region corresponding to the first attention region in the reconstructed picture, and encoding the residual data.

Abstract: Disclosed herein are an image processing method and apparatus for machine vision. The image processing method for machine vision includes generating an image with machine-unrecognizable distortion based on an input image, and encoding the image with machine-unrecognizable distortion.

Abstract: Disclosed herein is a method for compressing an image for machine vision, the method including detecting objects in an input image using an object detection network, generating a foreground image including bounding boxes corresponding to the objects and a background image, which is an image acquired by excluding the bounding boxes from the input image, encoding the foreground image and the background image, and decoding the encoded foreground image and the encoded background image.

Abstract: Disclosed herein is a method for generating a depth map from multi-view images, the method including receiving multi-view images and parameters thereof; generating an initial depth map, an initial normal map, and an initial reliability map for the multi-view images; performing selection of multiple initial vertices and triangulation based on the initial depth map and the initial normal map and generating depth values and normal values for the result of triangulation; performing regularization based on triangulation, selecting additional vertices in the initial depth map, and generating additional depth values and additional normal values based on the additional vertices and regularization; and refining the initial depth map into a refined depth map by replacing depth values and normal values having reliability below a preset level in the initial depth map and the initial normal map with the additional depth values and the additional normal values.

Abstract: Disclosed herein are a method, an apparatus and a storage medium for image encoding/decoding using a segmentation map. A feature vector for an image may be extracted using a segmentation map. The image may be encoded using the segmentation map and the feature vector. An output stream from an encoding apparatus may include a video stream and a feature stream. An input stream to a decoding apparatus may include a video stream and a feature stream. The image may be reconstructed using a reconstructed segmentation map and a reconstructed feature vector.

Abstract: Provided is a method of removing noise from a sparse depth map. The method includes setting, as nodes, sample points in a sparse depth map including depth information at a certain position of an image plane, constructing a mesh by connecting the set nodes, and removing noise from the meshed node, based on connection information between the meshed node and neighboring nodes adjacent thereto and color information about a color image corresponding to the sparse depth map. A depth value is allocated to each of the sample points.

Abstract: A method of encoding and reconstructing a depth image using color image information includes segmenting a color image and a depth image corresponding to the color image into block units, selecting a position of a pixel sampling value from each block of the depth image using the color image information, and encoding the pixel sampling value of the selected position as depth information.

Abstract: Provided is a method of removing noise from a sparse depth map. The method includes setting, as nodes, sample points in a sparse depth map including depth information at a certain position of an image plane, constructing a mesh by connecting the set nodes, and removing noise from the meshed node, based on connection information between the meshed node and neighboring nodes adjacent thereto and color information about a color image corresponding to the sparse depth map. A depth value is allocated to each of the sample points.

Abstract: An apparatus for multi-view stereo includes: an initial dense depth map generator to generate an initial dense depth map based on color information and mesh information from a sparse depth map; and a dense depth map improver to regenerate a dense depth map from a sparse depth map where points are added to the sparse depth map.

Abstract: A method of encoding and reconstructing a depth image using color image information includes segmenting a color image and a depth image corresponding to the color image into block units, selecting a position of a pixel sampling value from each block of the depth image using the color image information, and encoding the pixel sampling value of the selected position as depth information.