System and method for video processing
View Patent ↗Method and apparatus for decimating or sub-sampling image data which uses fixed delay intervals to accumulate a weighted sum of input samples. The accumulated sum is output at selected intervals, the interval determining the degree of decimation. The apparatus can take the form of a Finite Impulse Response (FIR) decimation filter. Decimation can be horizontal, vertical or temporal, and the decimation factor can be made non-integral.
1. A finite impulse response (FIR) decimation filter for sub-sampling an input image to output a sub-sampled image, said filter comprising:
a plurality of filter taps;
a fixed number of coefficient multipliers connected respectively in parallel to said filter taps, each multiplier producing a respective weighted contribution from said input; and
a fixed-length chain of serially-connected accumulators with the respective multiplier outputs connected to respective ones of said accumulators and where the accumulated values are passed along the chain at the decimated output sample rate,
wherein the decimation factor can be varied by changing the number of times values are accumulated in said accumulators.
2. A filter according to claim 1 in which the decimation factor is not an integer.
3. A filter according to claim 1 , where the decimation factor is changed by changing the number of times said weighted contributions from an input sample are accumulated in said serially connected, fixed length chain of accumulators.
4. A finite impulse response (FIR) decimation filter for sub-sampling an input image to output a sub-sampled image, said filter comprising:
a plurality of filter taps having fixed delays between the filter taps, and a recirculating storage loop arranged in stages, in which a weighted sum of input samples being a sum of weighted input samples is accumulated in said recirculating storage loop so that data is transferred from each respective stage to the next at the input sample rate and decimated samples are taken from one of the stages at output sample rate,
wherein the time for data to complete one circuit of the loop is less than a number of input sample periods equal to the decimation ratio, and
wherein the decimation factor can be varied by changing the number of times values are accumulated in said accumulators.
5. A filter according to claim 4 in which the decimation factor is not an integer.