Abstract: Techniques are disclosed relating to modifying an automatically predicted adjustment. In one embodiment, the automatically predicted adjustment may be adjusted, for example, based on a rule. The automatically predicted adjustment may be based on a machine learning prediction. A new image may be globally adjusted based on the modified automatically predicted adjustment.

Abstract: Techniques are disclosed relating to generating generic labels, translating generic labels to image pipeline-specific labels, and automatically adjusting images. In one embodiment, generic labels may be generated. Generic algorithm parameters may be generated based on training a regression algorithm with the generic labels. The generic labels may be translated to pipeline-specific labels, which may be usable to automatically adjust an image.

Abstract: Methods, systems, and apparatus, including computer program products, feature receiving user input defining a sample of pixels from an image, the image being defined by a raster of pixels. While receiving the user input, the following actions are performed one or more times: pixels are coherently classified in the raster of pixels as being foreground or background based on the sample of pixels; and a rendering of the image is updated on a display to depict classified foreground pixels and background pixels as the sample is being defined.

Abstract: Methods and apparatus for describing a projection model, used by a panoramic image stitching module to generate panoramic images and for communicating the projection model to other processes. A post-processing module may access and use the projection model provided by the panoramic image stitching module to perform one or more post-processing methods on the panoramic image, rather than requiring the user to input the projection model via a user interface or requiring the post-processing module to estimate the projection model according to a mathematical analysis of the panoramic image.

Abstract: Systems, methods, and apparatuses, including computer program products, are provided for re-layout of composite images. In some implementations, a method includes identifying geometric transformations corresponding to multiple images from a collection of images, where a geometric transformation reorients a corresponding image in relation to a common reference frame when applied and identifying a reference image for the multiple images in the collection of images. The method also includes determining overlapping image regions for the multiple images starting from the reference image, the determining based on the identified geometric transformations, determining additional transformations of a specified type for the multiple images based on the overlapping image regions, where an additional transformation lays out a corresponding image in relation to the reference image when applied, and making the additional transformations available for further processing and output with respect to the collection of images.

Abstract: Method and apparatus for determining sensor format factors from image metadata in which one of multiple techniques is applied to determine a sensor format factor for a camera from information in the metadata. Information from the metadata may be used to identify which of the multiple techniques to use. For example, the camera make and model may be used to determine a particular technique to use. As another example, the presence or absence of particular attributes or values for particular attributes may be used in determining a particular technique. In one embodiment, a profile database may be searched according to information in the metadata to determine if a sensor format factor for the camera is in the database. If the sensor format factor for the camera is not in the database, other information in the metadata may be examined to determine a particular technique from among the plurality of techniques.

Abstract: Methods and apparatus for processing collections of images are described in which metadata from a set of images may be used in directing a multi-image processing workflow. One or more output images may be rendered from a set of input images, with each output image being some combination of two or more of the input images. To render an output image, a workflow including one or more processing steps may be applied to the images. Metadata may be used in directing and performing the workflow. For example, metadata may be used in determining a particular workflow for a set of images. As another example, metadata may be used to sort a collection of images into multiple categories for automated workflow processing. As yet another example, metadata may be used to retrieve information stored in a profile database that may be used in processing the images.

Abstract: Techniques are disclosed relating to modifying an automatically predicted adjustment. In one embodiment, the automatically predicted adjustment may be adjusted, for example, based on a rule. The automatically predicted adjustment may be based on a machine learning prediction. A new image may be globally adjusted based on the modified automatically predicted adjustment.

Abstract: Methods and apparatus for manipulating digital images. A warping module is described that enables the manipulation of a surface by selectively deforming portions of the surface while maintaining local rigidity. The user may position multiple control points on a surface to constrain deformation. The user may specify multiple properties (e.g., translation, rotation, depth, and scale) at each control point. A mesh may be overlaid on the surface. The warping module may perform an initialization in which the properties are propagated other vertices in the mesh to generate an initial deformed mesh. The warping module may then perform an iterative optimization operation on the deformed mesh to improve the deformation while retaining local rigidity. Thus, instead of moving every pixel in the surface, the warping module moves or adjusts coordinates of the vertices of the mesh. The surface is then deformed according to the deformed mesh.

Abstract: Methods, apparatus, and computer-readable storage media for tone mapping High Dynamic Range (HDR) images. An input HDR image is separated into luminance and color. Luminance is processed to obtain a base layer and a detail layer. The base layer is compressed according to a non-linear remapping function to reduce the dynamic range, and the detail layer is adjusted. The layers are combined to generate output luminance, and the output luminance and color are combined to generate an output image. A base layer compression technique may be used that analyzes the details and compresses the base layer accordingly to provide space at the top of the intensity scale where the details are displayed to thus generate output images that are visually better than images generated using conventional techniques. User interface elements may be provided via which a user may control one or more parameters of the tone mapping method.

Abstract: Methods and apparatus for retargeting and prioritized interpolation of lens profiles. A lens profile file may include a set of lens sub-profiles. The camera body and/or settings described in the file may not exactly match that of camera body and/or settings used to capture a target image. A sub-profile processing module may perform a prioritized sub-profile sorting and interpolation method to generate an interpolated sub-profile that may be applied to the target image to correct aberrations including, but not limited to, geometric distortion, lateral chromatic aberration, and vignette. Thus, models generated for a reference camera at a variety of settings may be applied to a target image captured with the same type of lens but with a different camera and/or with different settings that are not exactly modeled in the lens profile file.

Abstract: A system and method for a blur brush performing adaptive bilateral filtering is disclosed. The method may include receiving user input selecting an area of an image to be filtered, such as by pointing to the image area using the blur brush. The selected image may comprise an edge and a plurality of pixels. The method may operate to the blur brush identifying the edge in the selected image area. The method may operate to apply a filter tool (e.g., a bilateral filter) to the selected image area, while preserving the edge. The methods may be implemented by program instructions executing in parallel on CPU(s) or GPUs.

Abstract: Methods and apparatus for soft edge masking. A soft edge masking technique may be provided via which, starting from an initial, potentially very rough and approximate border selection mask, the user may selectively apply brush strokes to areas of an image to selectively improve the border region of the mask, thus providing softness details in border regions which contain soft objects such as hair and fur. A stroke may be an additive stroke indicating a particular region in which detail from an original image is to be added to a composite image, or a subtractive stroke indicating a particular region in which detail is to be removed from the composite image. The stroke may also indicate a strength parameter value that may be used to indicate an amount of bias to be used in opacity calculations for the affected pixels.

Abstract: Systems, methods, and computer-readable storage media for automatically detecting skin tones in an input image are disclosed. An initial skin tone mask for the image may be created dependent on a general skin tone model. An upper threshold and/or a lower threshold may be applied to the initial skin tone mask to identify pixels most likely to be skin pixels and least likely to be skin pixels, respectively. These pixels may be used to produce an image-specific skin tone model, including one or more truncated Gaussian models for skin pixels and/or non-skin pixels defined in a three-dimensional color space. The image-specific skin tone model may be applied to the image to generate a final skin tone mask. Skin tone detection may be automatically or selectively performed in conjunction with image editing or image feature identification operations to target or exclude skin pixels or non-skin pixels during execution of the operations.

Abstract: Various embodiments of methods and apparatus for motion deblurring in text images are disclosed. In one embodiment, a threshold-based text prediction for a blurred image is generated. A point spread function for the blurred image is estimated. A result of the threshold-based text prediction function is deconvolved based on the point spread function. The generating, estimating, and deconvolving are iterated at a plurality of scales, and a final deconvolution of a result of the iteratively deconvolving is executed.

Abstract: Various embodiments of methods and apparatus for motion deblurring are disclosed. In one embodiment, an estimate of a latent image of a blurred image at a current scale from an estimate of a latent image at a previous coarse scale is generated using an upsampling super-resolution function, and a blur kernel is estimated based on the estimate of the latent image and the blurred image; and are repeated from a course to fine scale. A final image estimate is generated. The generating the final image estimate includes performing a deconvolution of the latent image using the blur kernel and the blurred image.

Abstract: A method, system, and computer-readable storage medium are disclosed for generating a mask based on a color model. In one embodiment, a selection of at least one color in an input image is received. A color model may be determined based on the selection of the at least one color, wherein the color model comprises one or more truncated Gaussian functions. A mask may be generated based on the color model. The mask may comprise a respective value indicating a degree of membership in the mask for each of the plurality of pixels in the input image, wherein the degree of membership in the mask is a function of a similarity in color to the selection.

Abstract: Methods and systems for image upscaling are disclosed. In one embodiment, a low frequency band image intermediate is obtained from an input image. The input image is upsampled by a scale factor to obtain an upsampled image intermediate. A result image is estimated based at least in part on the upsampled image intermediate, the low frequency band image intermediate, and the input image, wherein the input image is of a smaller scale than the result image.

Abstract: Methods and systems for denoising and artifact removal in image upscaling are disclosed. In one embodiment, a low frequency band image intermediate is obtained from an input image. An upsampled image intermediate is obtained from the input image by upsampling. A result image is estimated, based at least in part on the upsampled image intermediate, the low frequency band image intermediate, and the input image. The input image is of a smaller scale than the result image. The estimating the result image further includes eliminating from the result image noise that is present in the input image.

Abstract: Methods and systems for a regression-based learning model in image upscaling are disclosed. In one embodiment, a set of image patch pairs for each of a set of images is generated. Each of the image patch pairs contains a natural image and a corresponding downscaled lower-resolution image. A regression model based at least in part on the set of image patch pairs is defined. The regression model represents a gradient of a function of the downscaled lower-resolution image. An image is upscaled based at least in part on the regression model.