Microsoft Technology Licensing, LLC

Matthew Uyttendaele

Richard S Szeliski

Sing Bing Kang

Simon Winder

Charles Lawrence Zitnick, Iii

Michael F Cohen

Kentaro Toyama

Johannes P Kopf

Bernhard Schoelkopf

Drew Edward Steedly

William Crow

John R Douceur

Steven Mark Drucker

Ross Garrett Cutler

Desney S Tan

Eric H Rudolph

Patrick Markus Baudisch

Daniel Chaim Robbins

Antonio Criminisi

Ying-Qing Xu

Qifa Ke

Simon John Baker

Anitha Kannan

Nuria M Oliver

Andrew L Bliss

Eric Paul Bennett

Daniel Lischinski

Eric K Ringger

Krishnan Ramnath

Ramaswamy Hariharan

Qing Luan

Dahua Lin

Ronald Keith Logan

Charles L Zitnick

Chris Pal

R Alex Colburn

Juliet Anne Bernstein

Ashley Michelle Eden

Charles Zitnick Iii

An image in a web page may be annotated after deriving information about an image when the image may be displayed on multiple web pages. The web pages that show the image may be analyzed in light of each other to determine metadata about the image, then various additional content may be added to the image. The additional content may be hyperlinks to other webpages. The additional content may be displayed as annotations on top of the images and in other manners. Many embodiments may perform searching, analysis, and classification of images prior to the web page being served.

Image annotations on web pages

A process for compressing and decompressing non-keyframes in sequential sets of contemporaneous video frames making up multiple video streams where the video frames in a set depict substantially the same scene from different viewpoints. Each set of contemporaneous video frames has a plurality frames designated as keyframes with the remaining being non-keyframes. In one embodiment, the non-keyframes are compressed using a multi-directional spatial prediction technique. In another embodiment, the non-keyframes of each set of contemporaneous video frames are compressed using a combined chaining and spatial prediction compression technique. The spatial prediction compression technique employed can be a single direction technique where just one reference frame, and so one chain, is used to predict each non-keyframe, or it can be a multi-directional technique where two or more reference frames, and so chains, are used to predict each non-keyframe.

Compressing and decompressing multiple, layered, video streams employing multi-directional spatial encoding

A dynamic wide angle image viewing technique is presented which provides a way to view a wide-angle image while zooming between a wide angle view and a narrower angle view that employs both perspective and non-perspective projection models. In general, this involves first establishing the field of view for a view of the wide angle image that is to be displayed. The view is then rendered and displayed based on the established field of view, such that the projection transitions between a perspective projection associated with narrower angle views and a non-perspective projection (e.g., cylindrical, spherical or some other parameterization) associated with wider-angle views.

Dynamic viewing of wide angle images

Techniques are described for rendering annotations associated with an image. A view of an image maybe shown on a display, and different portions of the image are displayed and undisplayed in the view according to panning and/or zooming of the image within the view. The image may have annotations. An annotation may have a location in the image and may have associated renderable media. The location of the annotation relative to the view may change according to the panning and/or zooming. A strength of the annotation may be computed, the strength changing based the panning and/or zooming of the image. The media may be rendered according to the strength. Whether to render the media may be determined by comparing the strength to a threshold.

Rendering annotations for images

Multi-spline image blending technique embodiments are presented which generally employ a separate low-resolution offset field for every image region being blended, rather than a single (piecewise smooth) offset field for all the regions to produce a visually consistent blended image. Each of the individual offset fields is smoothly varying, and so is represented using a low-dimensional spline. A resulting linear system can be rapidly solved because it involves many fewer variables than the number of pixels being blended.

Image blending using multi-splines

Multi-pass image resampling technique embodiments are presented that employ a series of one-dimensional filtering, resampling, and shearing stages to achieve good efficiency while maintaining high visual fidelity. In one embodiment, high-quality (multi-tap) image filtering is used inside each one-dimensional resampling stage. Because each stage only uses one-dimensional filtering, the overall computation efficiency is very good and amenable to graphics processing unit (GPU) implementation using pixel shaders. This embodiment also upsamples the image before shearing steps in a direction orthogonal to the shearing to prevent aliasing, and then downsamples the image to its final size with high-quality low-pass filtering. This ensures that none of the stages causes excessive blurring or aliasing.

Multi-pass image resampling

A dynamic tone mapping technique is presented that produces a local tone map for a sub-image of a wide-angle, high dynamic range (HDR), which is used in rendering the sub-image for display. The technique generally involves first computing a global tone map of the wide-angle, HDR image in advance of rendering the sub-image. The global tone map is then used during rendering to compute a local tone map based on the average luminance and contrast of the pixels of the sub-image. In addition, the sub-image can be tone mapped as part of the rendering of a sequence of sub-images during a viewer-executed panning and/or zooming session. In this case, the local tone maps can be kept from changing too rapidly by adding a hysteresis feature to smooth out the intensity changes between successive sub-images.

Viewing wide angle images using dynamic tone mapping

A gigapixel image is generated from a set of images in raw format depicting different portions of a panoramic scene that has up to a full spherical field of view. Radiometric alignment of the images creates a set of images in radiance format. Geometric alignment of the radiance format images creates a set of true poses for the images in radiance format. A gigapixel image depicting the entire scene is assembled from the set of radiance format images and radiance format true poses for the images. The set of images in raw format is captured using a conventional digital camera, equipped with a telephoto lens, attached to a motorized head. The head is programmed to pan and tilt the camera in prescribed increments to individually capture the images at a plurality of exposures and with a prescribed overlap between images depicting adjacent portions of the scene.

Generating gigapixel images

An 'Oblique Image Stitcher' provides a technique for constructing a photorealistic oblique view from a set of input images representing a series of partially overlapping views of a scene. The Oblique Image Stitcher first projects each input image onto a geometric proxy of the scene and renders the images from a desired viewpoint. Once the images have been projected onto the geometric proxy, the rendered images are evaluated to identify optimum seams along which the various images are to be blended. Once the optimum seams are selected, the images are remapped relative to those seams by leaving the mapping unchanged at the seams and interpolating a smooth mapping between the seams. The remapped images are then composited to construct the final mosaiced oblique view of the scene. The result is a mosaic image constructed by warping the input images in a photorealistic manner which agrees at seams between images.

Image stitching using partially overlapping views of a scene

A 'Joint Bilateral Upsampler' uses a high-resolution input signal to guide the interpolation of a low-resolution solution set (derived from a downsampled version of the input signal) from low-to high-resolution. The resulting high-resolution solution set is then saved or applied to the original input signal to produce a high-resolution output signal. The high-resolution solution set is close to what would be produced directly from the input signal without downsampling. However, since the high-resolution solution set is constructed in part from a downsampled version of the input signal, it is computed using significantly less computational overhead and memory than a solution set computed directly from a high-resolution signal. Consequently, the Joint Bilateral Upsampler is advantageous for use in near real-time operations, in applications where user wait times are important, and in systems where computational costs and available memory are limited.

Joint bilateral upsampling

A location history is a collection of locations over time for an object. A stay is a single instance of an object spending some time in one place, and a destination is any place where one or more objects have experienced a stay. Location histories are parsed using stays and destinations. In a described implementation, each location of a location history is recorded as a spatial position and a corresponding time at which the spatial position is acquired. Stays are extracted from a location history by analyzing locations thereof with regard to a temporal threshold and a spatial threshold. Specifically, two or more locations are considered a stay if they exceed a minimum stay duration and are within a maximum roaming distance. Each stay includes a location, a starting time, and an ending time. Destinations are produced from the extracted stays using a clustering operation and a predetermined scaling factor.

Parsing location histories

Oblique image stitching

A technique for image compositing which allows a user to select the best image of an object, such as for example a person, from a set of images interactively and see how it will be assembled into a final photomontage. A user can select a source image from the set of images as an initial composite image. A region, representing a set of pixels to be replaced, is chosen by the user in the composite image. A corresponding same region is reflected in one or more source images, one of which will be selected by the user for painting into the composite image. The technique optimizes the selection of pixels around the user-chosen region or regions for cut points that will be least likely to show seams where the source images are merged in the composite image.

Region selection for image compositing

A Bayesian two-color image demosaicer and method for processing a digital color image to demosaic the image in such a way as to reduce image artifacts. The method and system are an improvement on and an enhancement to previous demosaicing techniques. A preliminary demosaicing pass is performed on the image to assign each pixel a fully specified RGB triple color value. The final color value of pixel in the processed image is restricted to be a linear combination of two colors. Fully-specified RGB triple color values for each pixel in an image used to find two clusters represented favored two colors. The amount of contribution from these favored two colors on the final color value then is determined. The method and system also can process multiple images to improve the demosaicing results. When using multiple images, sampling can be performed at a finer resolution, known as super resolution.

Bayesian demosaicing using a two-color image

Growing community of inventors

Seattle, WA, United States of America

Matthew Uyttendaele