Coding and noise filtering an image sequence
The invention provides a method of and a device for coding ( 10* ) an image sequence (g(i,j,k)). The device ( 10* ) has a noise filter ( 12 ) for noise filtering ( 12 ) the image sequence (g(i,j,k)), and for regarding the noise filtering ( 12 ) as a rate-distortion optimization problem in order to adapt the response of the noise filter ( 12 ). In particular, a filter parameter set (C) is determined to adapt the response of the filter ( 12 ) in that the image sequence (g(i,j,k)) is encoded using an optimal bit-budget, which is the bit-budget used for compressing a noise-free image sequence, and the distortion for the given bit-budget is minimized. A solution of the rate-distortion problem is proposed that calculates ( 12 ) the filter parameter set (C) by a Lagrange multiplier method. Separate minimization is used to determine each parameter of the parameter set (C) independently. In a practical embodiment, a spatial adaptive weighted averaging (SAWA) filtering ( 23 ) is used to estimate the bit budget (R f* ) and to pre-filter the image sequence (g(i,j,k)).
1. A method of coding ( 10 , 10 *) an image sequence (g(i,j,k), the method comprising the steps of:
noise filtering ( 12 ) the image sequence (g(i,j,k); and
adapting ( 12 *) said noise filtering ( 12 ) in dependence on a number of bits to code a noise-free image sequence by calculating ( 12 *) a filter parameter set (C) of said noise filtering ( 12 ) by a Lagrange multiplier method.
2. A method as claimed in claim 1 , wherein the adapting step ( 12 *) comprises the steps of:
estimating ( 12 *) an optimal Lagrange multiplier by determining a maximum of a second derivative of a rate used in the adapting said noise filtering step.
3. A method as claimed in claim 1 , wherein the adapting step comprises the steps of:
spatial filtering ( 23 ) the image sequence; and
taking ( 17 ) a number of bits that is used to encode the spatially filtered image sequence as a rate used in the adapting said noise filtering step.
4. A device ( 10 , 10 *) for coding an image sequence (g(i,j,k), the device comprising:
means ( 12 ) for noise filtering the image sequence (g(i,j,k), and
means ( 12 *) for adapting a response of said means for noise filtering ( 12 ) in dependence on a number of bits to code a noise-free image sequence by calculating ( 12 *) a filter parameter set (C) of said noise filtering ( 12 ) by a Lagrange multiplier method.
5. A camera system comprising a camera ( 100 ) and a device ( 200 ) for coding an image sequence (g(i,j,k), the device comprising:
means ( 12 ) for noise filtering the image sequence (g(i,j,k); and
means ( 12 *) for adapting a response of said means for noise filtering ( 12 ) in dependence on a number of bits to code a noise-free image sequence by calculating ( 12 *) a filter parameter set (C) of said noise filtering ( 12 ) by a Lagrange multiplier method.