Abstract: A method for reducing chrominance entropy in a digital color image. This method is practiced in relation to pixels in the two chrominance channels in a luminance-chrominance-chrominance representation of the image. The method involves shifting the chrominance values of selected low-luminance or high-luminance pixels toward (or to) another chrominance value which will result in a net chrominance entropy reduction in the subject channel.

Abstract: A system encodes video by detecting the location of a facial region of a frame of the video. Sensitivity information is calculated for each of a plurality of locations within the video based upon the location of the facial region. The frame is encoded in manner that provides a substantially uniform apparent quality of the plurality of locations to the viewer when the viewer is observing the facial region of the video.

Abstract: To improve compression efficiency and decompressed image quality, images are decomposed into separate layers each containing a limited number of image element types (text, line art, or photographic). Each layer can then be compressed separately with a process that is optimal for the type of image element included in the layer. Images are decomposed into foreground, background, and mask layers. The value of a pixel in the mask layer designates whether the spatially corresponding pixel of the recomposed image will be selected from the spatially corresponding pixels of either the foreground or the background layer.

Abstract: Images may be decomposed into separate layers each containing a limited number of types of image element (text, line art, or photographic). Each layer can then be compressed separately with a process that is optimal for the type of image element included in the layer. Images are decomposed into foreground, background, and mask layers. A method is provided for generating the out-of-layer pixels and, as a result, generating a layer having a limited color palette that is to be compressed with a palette based lossless compression method. A plurality of neighboring pixels is identified for each pixel selected from the image. If the neighboring pixels are out-of-layer pixels, the spatially corresponding pixel in the decomposition layer is assigned a default value. If the neighboring pixels include in-layer pixels, the pixel of the decomposition layer spatially corresponding to the selected pixel is assigned the predominant color of the in-layer neighbors.

Abstract: To improve compression efficiency and decompressed image quality, images are decomposed into separate layers each containing a limited number of image element types (text, line art, or photographic). Each layer can then be compressed separately with a process that is optimal for the type of image element included in the layer. Images are decomposed into foreground, background, and mask layers. The value of a pixel in the mask layer designates whether the spatially corresponding pixel of the recomposed image will be selected from the spatially corresponding pixels of either the foreground or the background layer.

Abstract: A method for robust human face tracking in the presence of multiple facial images includes taking a frame from a color video sequence as a current input image; filtering the current input image to form a filtered image; estimating the locations and sizes of faces in the filtered image based on a projection histogram of the filtered image; estimating face motion in the filtered image; and outputting the location and size of the tracked faces within the filtered image.

Abstract: A system encodes video by detecting the location of a facial region of a frame of the video. Sensitivity information is calculated for each of a plurality of locations within the video based upon the location of the facial region. The frame is encoded in manner that provides a substantially uniform apparent quality of the plurality of locations to the viewer when the viewer is observing the facial region of the video.