|
Usage
Usage of the ORIP
Phase-Deploy IntroductionThis section contains some details about the input arguments. They are the same for each executable. Select -x for help or just run a program to see its syntax. Remarks on general usage-i <name> The name of the input file. The input should be a YUV 4:2:0 file, the user can specify the width and the height with the -w -h parameters. -o <name> The output file. For all algorithms, this is a file that contains the filtered (or transformed) samples; we append zeros for the U,V components. For all filtering and transform applications, the output is the processed Y frame in integer (32-bit) format. For the motion estimation algorithms, the output is the motion-compensated Y frames, in unsigned character (8-bit) format, appended with zeros for the U,V components. -b <X> This is the number of bitplanes you would like to process in total (X is between 1 and 8). For the conventional approach, this means that each input frame will be inserted from bitplane 7 (most-significant bitplane) down to (and including) bitplane (8 - X), and then the input will be processed. For the incremental approach, this means that execution (per frame) will terminate once all bitplanes up to (and including) (8 - X) have been incrementally inserted and processed. -t <X> This selects which case you would like to run for the convolution-related experiments. In the future, we plan to allow arbitrary kernels provided by the user. For the time being:
-p <X1,..,XN> This selects the input bitplane scanning pattern. This parameter only affects the incremental approach where the input image(s) are processed progressively from the most-significant bits to the least-significant bits. For example: 3,3,2 will read (and process) three, three and two bitplanes at a time (from the most-significant to the least-significant) and have terminating bitplanes n={5,2,0}, as described in the convolution experiments of our manuscript. Only the result of the final terminating bitplane (controlled by the -b <X>) is stored to the disk. Other patterns are admissible, e.g. 1,1,1,2,3 will read and process one, one, one, two, three bitplanes (going from the most-significant to the least-significant ones) and have terminating bitplanes n={7,6,5,3,0} . -l <X> This affects only the incremental approach. This parameter selects the coverage percentage of the last step in the scanning pattern. This option takes a float value between 0.0 and 1.0 and performs packing, processing and unpacking only at the X*100% of the input during the last step in the scanning pattern. -q <name> This is the input file for the user defined convolution or transform kernel. It is a Comma Separated Value file. The first row defines number of columns and number of rows, with "number of rows" rows with "number of columns" elements. An example file can be found inside the package. -z <X> This affects only the incremental motion estimation approach. This parameter selects the spiral distance limit (parameter W); the default value is 56.25% of the search range, rounded down to the nearest integer. The spiral size cannot be larger than the search range (as selected with the -s parameter). The larger the spiral size, the closer the results will match the full-search (non-incremental) motion estimation and the higher the required computation time. -x prints a short help message with the parameters and their short description. |