Hewlett-packard Company

Other

Gregory S Yovanof

Balas K Natarajan

Kontantinos Konstantinides

Alexander I Drukarev

A method of filtering an image to reduce noise. The image is divided into subimages or 'blocks'. An optimum value of a threshold parameter is determined by repeatedly applying a Singular Value Decomposition ('SVD') transform to one of the blocks for various values of the threshold parameter and then using a lossless compression algorithm to determine the compressed size of the various transforms. The optimum threshold parameter is determined by finding the maximum of the second derivative of the compressed size with respect to the log of the threshold parameter. The SVD transform is then applied to all the blocks using the optimum value of the threshold parameter, and the transformed blocks are then recombined to provide the filtered image.

Method of filtering images using image compressibility to determine

The present invention is a fully JPEG compliant two-pass approach that can compress an arbitrary image to a predetermined fixed size file. The compression coding device and method according to the present invention estimates an activity metric based on the average sum of the absolute value of the quantized DCT coefficients per block. Given the activity metric, a mathematical model relating the image activity to the JPEG Q-factor for a given value of the target compression ratio provides an estimated Q-factor value that yields the design target ratio. This mathematical model is developed during a calibration phase which is executed once off line for a given image capturing device. The fact that our activity metric is based on the quantized DCT coefficients allows for an efficient implementation of the presented coding method in either speed or memory bound systems.

Growing community of inventors

Sunnyvale, CA, United States of America

Gregory S Yovanof