Abstract: Techniques are described to generate improved training data for pixel labeling. To generate training data, objects are displayed in a user interface by a computing device, e.g., iteratively. The objects are taken from a structured object representation associated with a respective one of a plurality of images. The structured object representation defines a hierarchical relationship of the objects within the respective image. Inputs are then received that are originated through user interaction with the user interface. The inputs label respective ones of the iteratively displayed objects, e.g., as text, a graphical element, background, foreground, and so forth. A model is trained by the computing device using machine learning.

Abstract: Techniques are described to generate improved training data for pixel labeling. To generate training data, objects are displayed in a user interface by a computing device, e.g., iteratively. The objects are taken from a structured object representation associated with a respective one of a plurality of images. The structured object representation defines a hierarchical relationship of the objects within the respective image. Inputs are then received that are originated through user interaction with the user interface. The inputs label respective ones of the iteratively displayed objects, e.g., as text, a graphical element, background, foreground, and so forth. A model is trained by the computing device using machine learning.

Abstract: An image prior as a shared basis mixture model is described. In one or more implementations, a plurality of image patches are generated from one or more images. A shared basis mixture model is learned to model an image patch distribution of the plurality of image patches from the one or more images as part of a Gaussian mixture model. An image may then be reconstructed using the shared basis mixture model as an image prior.

Abstract: An image prior as a shared basis mixture model is described. In one or more implementations, a plurality of image patches are generated from one or more images. A shared basis mixture model is learned to model an image patch distribution of the plurality of image patches from the one or more images as part of a Gaussian mixture model. An image may then be reconstructed using the shared basis mixture model as an image prior.

Abstract: A logic arrangement (120), a storage medium (130), and a method for generating (FIG. 10b) a first digital image (430), are provided. In particular, a second digital image (410) can be generated using one or more brush strokes (200), and the second digital image can be modified based on particular data which is associated with the one or more brush strokes (200) so as to obtain the first digital image (430). Moreover, the first digital image (430) may have a perception of depth. For example, the particular data can include further data associated with a first relative height and a second relative height of the second digital image (410) at a plurality of locations within the second digital image (410). The first relative height of the second digital image (410) at a first location of the plurality of locations may be different than the second relative height of the second digital image (410) at a second location of the plurality of locations.

Abstract: A method and system receives a digital source image and brush size data. The source image is blurred to generate a digital reference image. The brush size data includes a first record corresponding to a first size of a brush and a second record corresponding to a second size of the brush. The first size is different from the second size. The method and system applies brush strokes, with the first record to be used for the brush, to a digital canvas image using the reference image. Then the brush strokes are applied, with the second record to be used for the brush, to the canvas image using the reference and working images. Thus, a final digital image having a hand-painted appearance is generated on the canvas image. Long curved brush strokes can also be used to generate the target image, which is aligned in a normal direction to image gradients. Furthermore the graphic artist may adjust parameters of the method and system according to the present invention to vary the style of painting.