Abstract: Techniques are described for real-time disparity upsampling for phase detection autofocus (PDAF) in digital imaging systems. For example, an imaging sensor array includes a number of physical PDAF sensors that provide a PDAF resolution that is a small percentage of the resolution of the imaging sensor itself. The PDAF resolution is upsampled without relying on adding physical PDAF sensors by generating simulated PDAF sensors at upsampling locations. For each upsampling location, embodiments can compute local disparity values and local contrast values for each of multiple neighboring physical PDAF sensors. An upsampled disparity value can be computed for the upsampling location by weighting the neighboring local disparity values based at least on the neighboring local contrast values. The upsampled disparity values can yield higher-resolution information for use in PDAF operations.

Abstract: Systems and methods for reducing noise in an image are provided. Noise is reduced in a luminance channel of the image using a first filtering procedure. Noise is reduced in a chrominance channel of the image using a second filtering procedure. The chrominance channel is decomposed into a plurality of frequency sub-bands, where each frequency sub-band of the plurality of frequency sub-bands represents the chrominance channel at a first resolution. The noise is further reduced in the chrominance channel using a third filtering procedure. The third filtering procedure is applied to each frequency sub-band of the plurality of frequency sub-bands. A lowest frequency sub-band of the plurality of frequency sub-bands is decomposed into a second plurality of frequency sub-bands after the third filtering procedure. Each frequency sub-band of the second plurality of frequency sub-bands represents the chrominance channel at a second resolution that is lower than the first resolution.

Abstract: An apparatus, method, and other embodiments associated with performing interpolations to compute gain values that correct for varying spatial intensity are described. In one embodiment, a method includes determining, by an apparatus that processes image data, a gain value for a pixel in the image data for which there is no gain value available in the apparatus, by interpolating related gain values associated with corners of a rectangle bounding the pixel, wherein the interpolating includes determining at least two partial coefficients by interpolating pairs of the related gain values. Noise is filtered from the image data using a noise threshold, and the noise threshold is modified by using the at least two partial coefficients. The method also applies the gain value to the pixel in the image data.

Abstract: A system and method of adaptive edge detection and noise reduction in an image where edge information is detected for each color component of each pixel, whether sensed or synthesized. In some embodiments, the filter applied to a selected non-edge pixel may be determined by the ultimate size of a region around the selected pixel, where the size of the region may be increased if a count of the non-edge pixels in the region is less than a threshold value.

Abstract: An apparatus and other embodiments associated with performing interpolations to compute gain values that correct for varying spatial intensity are described. In one embodiment, an apparatus includes interpolation logic configured to determine a gain value for a pixel in image data for which there is no gain value available in the apparatus. The interpolation logic is configured to determine the gain value by performing an interpolation of related gain values available in the apparatus. The apparatus also includes falloff correction logic configured to apply the gain value to the pixel in the image data.

Abstract: Apparatus, methods, and other embodiments associated with repairing defective pixels are described. In one embodiment, an apparatus includes pixel defect logic configured to retrieve pixel repair information associated with at least one defective sensor of an array of photo sensors. Pixel replication logic is configured to repair defective pixels in a stream of image pixels based, at least in part, on the pixel repair information.

Abstract: A system and method of adaptive edge detection and noise reduction in an image where edge information is detected for each color component of each pixel, whether sensed or synthesized. In some embodiments, the filter applied to a selected non-edge pixel may be determined by the ultimate size of a region around the selected pixel, where the size of the region may be increased if a count of the non-edge pixels in the region is less than a threshold value.

Abstract: A process of video capture, segmentation, tracking, and mapping to obtain position and time data for motion analysis. The moving object is captured in video in a series of frames at a known frame rate. The moving object is isolated from all other data in a frame and the position of the moving object is determined relative to stationary reference elements in the frame. However, the video captures a physical, three-dimensional space, in a two-dimensional image, causing distortions. This method associates all points in the quadrilateral with a rectangle defined by sides in real, physical space using a vanishing point algorithm. The preferred vanishing point algorithm maps a physical space to a quadrilateral defined by two sides that intersect at a first vanishing point, and two other sides that intersect at a second vanishing point. Single-point or three-point perspective may also be employed.

Abstract: Embodiments of a method of generating Huffman code length information are disclosed. In one such embodiment, a data structure is employed, although, of course, the invention is not limited in scope to the particular embodiments disclosed.

Abstract: A process of video capture, segmentation, tracking, and mapping to obtain position and time data for motion analysis. The moving object is captured in video in a series of frames at a known frame rate. The moving object is isolated from all other data in a frame and the position of the moving object is determined relative to stationary reference elements in the frame. However, the video captures a physical, three-dimensional space, in a two-dimensional image, causing distortions. This method associates all points in the quadrilateral with a rectangle defined by sides in real, physical space using a vanishing point algorithm. The preferred vanishing point algorithm maps a physical space to a quadrilateral defined by two sides that intersect at a first vanishing point, and two other sides that intersect at a second vanishing point. Single-point or three-point perspective may also be employed.

Abstract: Briefly, in accordance with one embodiment of the invention, a method of using hue to interpolate color pixel signals includes the following. For a particular pixel location in a subsampled color image, differences in hue are compared for two mutually orthogonal directions across the particular pixel location. A color signal value for that particular pixel location for a color plane other than the color plane of the pixel signal value in the subsampled color image of that location is computed. The computation includes relatively weighing the differences in hue values, the relative weights depending, at least in part, on the difference in hue value in one direction relative to the other.

Abstract: Embodiments of a method of generating Huffman code length information are disclosed. In one such embodiment, a data structure is employed, although, of course, the invention is not limited in scope to the particular embodiments disclosed.

Abstract: Embodiments of a method of generating Huffman code length information are disclosed. In one such embodiment, a data structure is employed, although, of course, the invention is not limited in scope to the particular embodiments disclosed.

Abstract: Embodiments of a method of performing Huffman decoding are disclosed. In one such embodiment, a data structure is employed, although, of course, the invention is not limited in scope to the particular embodiments disclosed.

Abstract: Briefly, in accordance with one embodiment of the invention, a method of inverse quantizing quantized signal samples of an image during image decompression includes the following. A process to transform the signal samples from a first domain to a second domain is applied. During the transform process, signal samples are filter, by first applying scaled filter coefficients to signal samples along the image in a first direction and then applying scaled filter coefficients to signal samples along the image in a second direction, so that at the completion of the transform process of the image, at least a selected portion of the transformed signal samples are inverse quantized. Many other embodiments in accordance with the invention are also described.

Abstract: A method includes capturing an optical image to form raw data indicative of the optical image and using a look-up table to transform the raw data into transformed data that indicates a second image. A white color balance of the second image is computed, and the values in the look-up table are modified based on the computed white color balance.

Abstract: An imaging system using a color shutter receives signals indicative of color information in a plurality of planes together with infrared information and a signal indicative of the infrared radiation alone. The infrared radiation signal may be subtracted from the signals associated with the color planes to develop signals free of infrared radiation effects for each of the desired color planes. In addition an infrared signal is available which may have useful applications in some situations. In this way, color information color bands may be obtained without the need an infrared filter and infrared information is retained for use in other applications if desired. In addition infrared and dark current information may be derived that can be used for dark current correction.

Abstract: A system of indexing discrete wavelet transform data, after entropy encoding, allows access to particular frames and sub-bands without decomposing all of the data to find the information of interest. In one embodiment a header may include the frame size in bytes followed by the sizes of each of the sub-bands in a predetermined order.

Abstract: Numerous embodiments for a method of performing Huffman decoding are disclosed.

Abstract: An imager with a four color mosaic pattern of red, green, blue, and infrared pass filters, where color component signals for a pixel are interpolated by averaging over nearest neighbor pixels.