Nvidia Corporation

Sony Corporation

Sony Electronics Inc

Elif Albuz

Shirish C Gadre

Shalini Gupta

Fang-Chuan Wu

Mateusz Jerzy Baranowski

Raman Subramanian

Tarik Arici

Robert Jan Schutten

Santanu Dutta

Gong-San Yu

Leonid Abraham Milenky

Ismail Oner Sebe

In one embodiment of the present invention, a graphics processing unit (GPU) is configured to detect an object in an image using a random forest classifier that includes multiple, identically structured decision trees. Notably, the application of each of the decision trees is independent of the application of the other decision trees. In operation, the GPU partitions the image into subsets of pixels, and associates an execution thread with each of the pixels in the subset of pixels. The GPU then causes each of the execution threads to apply the random forest classifier to the associated pixel, thereby determining a likelihood that the pixel corresponds to the object. Advantageously, such a distributed approach to object detection more fully leverages the parallel architecture of the parallel processing unit (PPU) than conventional approaches. In particular, the PPU performs object detection more efficiently using the random forest classifier than using a cascaded classifier.

Performing object detection operations via a graphics processing unit

In one embodiment of the present invention, a graphics processing unit (GPU) is configured to detect an object in an image using a random forest classifier that includes multiple, identically structured decision trees. Notably, the application of each of the decision trees is independent of the application of the other decision trees. In operation, the GPU partitions the image into subsets of pixels, and associates an execution thread with each of the pixels in the subset of pixels. The GPU then causes each of the execution threads to apply the random forest classifier to the associated pixel, thereby determining a likelihood that the pixel corresponds to the object. Advantageously, such a distributed approach to object detection more fully leverages the parallel architecture of the PPU than conventional approaches. In particular, the PPU performs object detection more efficiently using the random forest classifier than using a cascaded classifier.

Performing object detection operations via random forest classifier

A method for more efficiently detecting faces in images is disclosed. The integral image of an image may be calculated. The integral image may be sub-sampled to generate one or more sub-sampled integral images. A plurality of classifiers may be applied in one or more stages to regions of each sub-sampled integral image, where the application of the classifiers may produce classification data. The classification data may be used to determine if a face is associated with any of the regions of each sub-sampled integral image. The face determination results may be used to modify the original image such that, when rendered, the image is displayed with a graphical object identifying the face in the image. Accordingly, face detection processing efficiency may be increased by reducing the number of integral image calculations and processing localized data through application of classifiers to sub-sampled integral images.

Application of classifiers to sub-sampled integral images for detecting faces in images

A method includes projecting motion vectors describing a transformation from a previous video frame to a future video frame onto a plane between the previous video frame and the future video frame, detecting potential artifacts at the plane based on an intersection of a cover region and an uncover region on the plane, and analyzing a dissimilarity between a trial video frame and both the previous video frame and the future video frame. The trial video frame is generated between the previous video frame and the future video frame based on a frame rate conversion ratio derived from a source frame rate and a desired frame rate. The method also includes estimating reliability of the projected motion vectors based on the potential artifact detection and the dissimilarity analysis.

Reducing and correcting motion estimation artifacts during video frame rate conversion

A system, method, and computer program product are provided for refining motion vectors. In operation, a plurality of motion vectors associated with a current frame and a first resolution are created. Furthermore, the motion vectors are refined utilizing information including at least one of first information describing motion vectors associated with a previous frame and second information describing motion vectors associated with the current frame and a second resolution.

System, method, and computer program product for refining motion vectors

Methods and apparatus for implementing partial and start-over threads in a kernel of an operating system are disclosed. In a computing system having at least one CPU, registers for executing threads, and memory, a method for executing a partial thread includes executing a first thread in the registers, initiating an interrupt event that signals the activation of another thread, saving a first portion of registers to a memory location, wherein a second portion of the registers is unsaved, the first portion including information associated with the first thread, executing in the registers, the another thread, and conducting further computing system operations. Additionally, conducting further computing system operations can include restoring the saved first portion of registers from the memory location to the registers and resuming execution of the first thread. Also, another embodiment incorporates the foregoing method into a computer device.

Partial and start-over threads in embedded real-time kernel

A dynamic allocation of available ASV buffer memory space is performed on each pack in a DVD-A bitstream one pack at a time. Concurrently, an ASV buffer table is updated for each type of data pack currently being processed. The ASV buffer table includes pointers corresponding to the various fields that form a particular ASV frame. In this way, only that memory that is required to store a particular ASV frame is used thereby allowing the ASV buffer memory to be configured on the fly in such a manner as to efficiently store the required ASV frame data. When a particular ASV frame is to be displayed, or otherwise processed, the ASV buffer table is accessed, and the particular pointers for a specific ASV frame are looked up and used to access the desired ASV frame.

Method and apparatus for efficiently allocating memory in audio still video (ASV) applications

An apparatus and method for providing enhanced performance for multi-processor multimedia chips. In one embodiment, the present invention is comprised of a data and communication apparatus coupled with the multimedia system in which the multi-processor multimedia chips are disposed. The present invention is comprised of a data memory to retrievably store data. The present invention is further comprised of an instruction memory to retrievably store instructions. The present invention is also comprised of an incoming buffer which permits transfer of data into the data and communication apparatus and provides fast access to streaming data. The present invention is additionally comprised of an outgoing buffer which monitors and permits transfer of data out of the data and communication apparatus.

Multifunctional I/O organizer unit for multiprocessor multimedia chips

When switching between a DVD-video mode and a DVD-audio mode in a DVD-A/V player, a current video frame is stored in a current display buffer portion of the memory during the DVD-video mode. The DVD-A/V player is paused in the DVD-video mode and set in the DVD-audio mode. If it is determined that the current display buffer portion of the memory is a reserved display buffer portion of the memory, then the current video frame is copied to a reconstructed display buffer portion of the memory. At least the current display portion of the memory is designated as an ASV buffer and a frame buffer management scheme is changed so as to preserve the ASV buffer.

Growing community of inventors

Sunnyvale, CA, United States of America

Elif Albuz