发明名称 Image quality control techniques
摘要 A technique for controlling the quality of one or more compressed images. The technique allows, for example, the selection of a target quality metric(s) and the compression of the image(s) such the compressed image(s) meets the metric(s). Alternatively, a target quality metric can be specified, and the image(s) compressed using parameters estimated to achieve the target quality. Optionally, the quality metric can also be made available to, for example, a user on an image processing system. The quality metrics can be, for example, for one or more layers, one or more images and/or one or more image sequences.
申请公布号 US9076190(B2) 申请公布日期 2015.07.07
申请号 US201414171476 申请日期 2014.02.03
申请人 AWARE, INC. 发明人 Tzannes Alexis Paul;Gut Ron Abraham
分类号 G06K9/36;G06K9/46;G06T9/00;H04N19/147;H04N19/172;H04N19/63;H04N19/61;H04N19/154;H04N19/162;H04N19/179;H04N19/115;H04N19/124;H04N19/146 主分类号 G06K9/36
代理机构 Sheridan Ross, PC 代理人 Vick Jason H.;Sheridan Ross, PC
主权项 1. A method to control a quality of one or more of an image, a sequence of images, a video sequence, a 3-dimensional data set and a layer of an image comprising: receiving the one or more of the image, the sequence of images, the video sequence, the 3-dimensional data set and the layer of the image; receiving an indication of a target image quality for a reconstructed version of a to be compressed version of the one or more of the image, the sequence of images, the video sequence, the 3-dimensional data set and the layer of the image; commencing compression of the one or more of the image, the sequence of images, the video sequence, the 3-dimensional data set and the layer of the image, the compression including performing a wavelet transform and wavelet coefficient coding; estimating, using one or more of Laplacian, generalized Gaussian, statistical and empirical modeling, distortion caused by one or more of quantization distortion and truncation distortion, wherein the estimating of distortion is performed without measuring the distortion; estimating, using one or more modeling techniques, a quality of the reconstructed version of the one or more of the image, the sequence of images, the video sequence, the 3-dimensional data set and the layer of the image based on the estimated distortion; automatically comparing, by a processor, the estimated quality to the target image quality; and: when the target image quality has been met, outputting the compressed version of the one or more of the image, the sequence of images, the video sequence, the 3-dimensional data set and the layer of an image that at least meets the target image quality, andwhen the target image quality has not been met, adjusting one or more of quantization tables and truncation points.
地址 Bedford MA US