Lucent Technologies Inc.

Paul S Giangrossi

Marc J Beacken

Robert L Doran

Paul R Gloudemans

Elizabeth A Seip

A composite image generator for generating a multi-image composite from a plurality of individual compressed images is disclosed. Multi-image composites are generated from static images or from a video stream with significantly reduced computation, reduced latency for video processing and reduced storage requirements. The individual images may be compressed with a JPEG image encoder or another suitable intraframe image encoder, such as those implementing the MPEG-II standard. In the illustrative embodiment, four (4) individual images of equal size, compressed using the JPEG standard, are combined into a multi-image composite with a single individual image in each quadrant of the composite image. Multi-image composites are created without fully decoding each individual compressed image, by partially decoding the individual compressed images and only fully decoding a small portion of each image. Multi-image composites are created from N individual compressed images by initially decoding the data of the N compressed images only enough to identify where each row of 8.times.8 blocks end, so that the image data may be reordered in an appropriate manner. As each 8.times.8 block is detected, the input image data is copied to the output image data to create the multi-image composite. The desired composite configuration determines the order of processing the N input images. Since each codeword may encode more than one coefficient, and each codeword may be of variable length, detection tables preferably indicate the number of bits encapsulated in each codeword and the number of coefficients represented by these bits. As the input image data is evaluated, the number of coefficients may be counted until all of the coefficients in a block have been processed, or until an end-of-block codeword is detected. The DC difference values at the border of each individual image must be recalculated to reflect the reordering of the pixels, since the DC coefficient in each block is encoded relative to the DC coefficient from the previous block, which will be different for certain blocks in the reordered multi-image composite image.

Growing community of inventors

New York, NY, United States of America

Paul S Giangrossi