Abstract: Embodiments of the present disclosure relate to a configurable convolution engine that receives configuration information to perform convolution or its variant operations on streaming input data of various formats. To process streaming input data, input data of multiple channels are received and stored in an input buffer circuit in an interleaved manner. Data values of the interleaved input data are retrieved and forwarded to multiplier circuits where multiplication with a corresponding filter element of a kernel is performed. Varying number of kernels with different sizes and sparsity can also be used for the convolution operations.

Abstract: Embodiments relate to an architecture of a vision pipe included in an image signal processor. The architecture includes a front-end portion that includes a pair of image signal pipelines that generate an updated luminance image data. A back-end portion of the vision pipe architecture receives the updated luminance images from the front-end portion and performs, in parallel, scaling and various computer vision operations on the updated luminance image data. The back-end portion may repeatedly perform this parallel operation of computer vision operations on successively scaled luminance images to generate a pyramid image.

Abstract: Embodiments of the present disclosure relate to highlight recovery of a high-resolution image using a single low-resolution image captured at a lower exposure. An example apparatus includes a hue target circuit that receives an input image at a high-resolution including at least one pixel with a clipped color channel. For example, the input image is a Blue sky image with a pixel having clipped Blue channel. The hue target circuit also receives a set of candidate hue maps having a pixel resolution lower than the high-resolution of the input image. The hue target circuit generates a target hue value for the at least one pixel using the pixel information of the set of candidate hue maps. The apparatus also includes a hue recovery circuit that generates a recovered version of the input image by adjusting hue information of the clipped color channel based on the generated target hue.

Abstract: Embodiments relate to a histogram-of-oriented gradients (HOG) module. The HOG module is implemented in hardware rather than software. The HOG module applies an algorithm to an image to identify gradient orientation in localized portions of the image. The HOG module creates a histogram-of orientation gradients based on the identified gradient orientations.

Abstract: Embodiments of the present disclosure relate to a configurable convolution engine that receives configuration information to perform convolution or its variant operations on streaming input data of various formats. To process streaming input data, input data of multiple channels are received and stored in an input buffer circuit in an interleaved manner. Data values of the interleaved input data are retrieved and forwarded to multiplier circuits where multiplication with a corresponding filter element of a kernel is performed. Varying number of kernels with different sizes and sparsity can also be used for the convolution operations.

Abstract: The present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. In particular, the present disclosure provides low power frameworks for processing, compressing, and transmitting images at a mobile image capture device. One example low power framework includes a scene analyzer that analyzes a scene depicted by a first image and determines whether to store the first image in a non-volatile memory or to discard the first image from a temporary image buffer without storing the first image in the non-volatile memory.

Abstract: An image signal processor may include a sensor interface that includes a pixel defect preprocessing (PDP) component that performs an initial adjustment of pixel values for patterned defect pixels in raw pixel data captured by an image sensor. To adjust a patterned defect pixel, the PDP component may apply an interpolation technique to values in a gain lookup table according to the pixel's location in the image frame to determine the gain value for the pixel, and then apply the gain value to the pixel. The PDP component may provide the raw pixel data with the adjusted patterned defect pixels to two or more other modules for additional processing. The other modules may include an image processing pipeline that may detect other defective pixels in the raw pixel data and correct the patterned defect pixels and the other defective pixels, for example using a weighted combination of neighboring pixels.

Abstract: An input rescale module that performs cross-color correlated downscaling of sensor data in the horizontal and vertical dimensions. The module may perform a first-pass demosaic of sensor data, apply horizontal and vertical scalers to resample and downsize the data in the horizontal and vertical dimensions, and then remosaic the data to provide horizontally and vertically downscaled sensor data as output for additional image processing. The module may, for example, act as a front end scaler for an image signal processor (ISP). The demosaic performed by the module may be a relatively simple demosaic, for example a demosaic function that works on 3×3 blocks of pixels. The front end of module may receive and process sensor data at two pixels per clock (ppc); the horizontal filter component reduces the sensor data down to one ppc for downstream components of the input rescale module and for the ISP pipeline.

Abstract: An image processing pipeline may perform noise filtering and image sharpening utilizing common spatial support. A noise filter may perform a spatial noise filtering technique to determine a filtered value of a given pixel based on spatial support obtained from line buffers. Sharpening may also be performed to generate a sharpened value of the given pixel based on spatial support obtained from the same line buffers. A filtered and sharpened version of the pixel may be generated by combining the filtered value of the given pixel with the sharpened value of the given pixel. In at least some embodiments, the noise filter performs spatial noise filtering and image sharpening on a luminance value of the given pixel, when the given pixel is received in a luminance-chrominance encoding.

Abstract: Systems and methods for local tone mapping are provided. In one example, an electronic device includes an electronic display, an imaging device, and an image signal processor. The electronic display may display images of a first bit depth, and the imaging device may include an image sensor that obtains image data of a higher bit depth than the first bit depth. The image signal processor may process the image data, and may include local tone mapping logic that may apply a spatially varying local tone curve to a pixel of the image data to preserve local contrast when displayed on the display. The local tone mapping logic may smooth the local tone curve applied to the intensity difference between the pixel and another nearby pixel exceeds a threshold.

Abstract: Systems and methods for correcting intensity drop-offs due to geometric properties of lenses are provided. In one example, a method includes receiving an input pixel of the image data, the image data acquired using an image sensor. A color component of the input pixel is determined. A gain grid is determined by pointing to the gain grid in external memory. Each of the plurality of grid points is associated with a lens shading gain selected based upon the color of the input pixel. A nearest set of grid points that enclose the input pixel is identified. Further, a lens shading gain is determined by interpolating the lens shading gains associated with each of the set of grid points and is applied to the input pixel.

Abstract: Systems and methods for reducing chrominance (chroma) noise in image data are provided. In one example of such a method, image data in YCC format may be received into logic of an image signal processor. Using the logic, noise may be filtered from a first chrominance component or a second chrominance component, or both, of the image data, using a sparse filter and a noise threshold. The noise threshold may be determined based at least in part on two of the components of the YCC image data.

Abstract: Techniques to detect subject and camera motion in a set of consecutively captured image frames are disclosed. More particularly, techniques disclosed herein temporally track two sets of downscaled images to detect motion. One set may contain higher resolution and the other set lower resolution of the same images. For each set, a coefficient of variation may be computed across the set of images for each sample in the downscaled image to detect motion and generate a change mask. The information in the change mask can be used for various applications, including determining how to capture a next image in the sequence.

Abstract: In an embodiment, an electronic device may be configured to capture still frames during video capture, but may capture the still frames in the 4×3 aspect ratio and at higher resolution than the 16×9 aspect ratio video frames. The device may interleave high resolution, 4×3 frames and lower resolution 16×9 frames in the video sequence, and may capture the nearest higher resolution, 4×3 frame when the user indicates the capture of a still frame. Alternatively, the device may display 16×9 frames in the video sequence, and then expand to 4×3 frames when a shutter button is pressed. The device may capture the still frame and return to the 16×9 video frames responsive to a release of the shutter button.

Abstract: An image processing pipeline may account for clipped pixels in auto focus statistics. Generating auto focus statistics may include evaluating a neighborhood of pixels with respect to a given pixel in a stream of pixels for an image. If a clipped pixel is identified within the neighborhood of pixels then the evaluation of the given pixel may be excluded from an auto focus statistic. The image processing pipeline may also provide auto focus statistics that do not exclude clipped pixels. A luminance edge detection value may, in some embodiments, be generated by applying an IIR filter to the given pixel in a stream of pixels to band-pass filter the given pixel before including the band-pass filtered pixel in the generation of the luminance edge detection value.

Abstract: An integrated stacked and/or abutted sensor, memory and processing hardware camera solution is described. The sensor is to receive light from an image and generate electronic pixels from the light. The processing hardware is to process the electronic pixels to: a) recognize a scene from the image in a lower quality image mode; b) trigger actions by the camera solution in response to the recognition of the scene, the actions including: i) transitioning the camera solution from the lower quality image mode to a higher quality image mode to capture a higher quality version of the image; and, ii) forwarding from the camera solution important imagery and not forwarding from the camera solution unimportant imagery.

Abstract: Systems and methods for automatic lens flare compensation may include a non-uniformity detector configured to operate on pixel data for an image in an image sensor color pattern. The non-uniformity detector may detect a non-uniformity in the pixel data in a color channel of the image sensor color pattern. The non-uniformity detector may generate output including location and magnitude values of the non-uniformity. A lens flare detector may determine, based at least on the location and magnitude values, whether the output of the non-uniformity detector corresponds to a lens flare in the image. In some embodiments, the lens flare detector may generate, in response to determining that the output corresponds to the lens flare, a representative map of the lens flare. A lens flare corrector may determine one or more pixel data correction values corresponding to the lens flare and apply the pixel data correction values to the pixel data.

Abstract: An image signal processor may include a pixel defect correction component that tracks defect history for frames captured by an image sensor and applies the history when identifying and correcting defective pixels in a frame. The component maintains a defect pixel location table that includes a defect confidence value for pixels of the image sensor. The component identifies defective pixels in a frame, for example by comparing each pixel's value to the values of its neighbor pixels. If a pixel is detected as defective, its defect confidence value may be incremented. Otherwise, the value may be decremented. If a pixel's defect confidence value is over a defect confidence threshold, the pixel is considered defective and thus may be corrected. If a pixel's defect confidence value is under the threshold, the pixel is considered not defective and thus may not be corrected even if the pixel was detected as defective.

Abstract: In an embodiment, an electronic device may be configured to capture still frames during video capture, but may capture the still frames in the 4×3 aspect ratio and at higher resolution than the 16×9 aspect ratio video frames. The device may interleave high resolution, 4×3 frames and lower resolution 16×9 frames in the video sequence, and may capture the nearest higher resolution, 4×3 frame when the user indicates the capture of a still frame. Alternatively, the device may display 16×9 frames in the video sequence, and then expand to 4×3 frames when a shutter button is pressed. The device may capture the still frame and return to the 16×9 video frames responsive to a release of the shutter button.

Abstract: An image processing pipeline may perform noise filtering and image sharpening utilizing common spatial support. A noise filter may perform a spatial noise filtering technique to determine a filtered value of a given pixel based on spatial support obtained from line buffers. Sharpening may also be performed to generate a sharpened value of the given pixel based on spatial support obtained from the same line buffers. A filtered and sharpened version of the pixel may be generated by combining the filtered value of the given pixel with the sharpened value of the given pixel. In at least some embodiments, the noise filter performs spatial noise filtering and image sharpening on a luminance value of the given pixel, when the given pixel is received in a luminance-chrominance encoding.