Changeset 26007 for branches/pap/psModules/src/imcombine/pmStack.h

Timestamp:

Nov 2, 2009, 5:08:31 PM (17 years ago)

Author:

Paul Price

Message:

Serious rework of pmStackCombine for clarity and some additional functionality. Split the combinePixels function into two operating modes (test and combine) to make clear what we're doing; pmStackCombine could be split into two separate functions (e.g., pmStackTest and pmStackCombine) in the future, since the testing part doesn't need to produce any output image, but this involves moving a lot of stuff around in ppStack, so not doing it yet. Verified that variance scaling is correct, and softening of variance is correct. Big improvement in minimising loss of good pixels comes from throwing out only the most variant pixel on each rejection pass (consequently changed iterations parameter to iterations per input); this involved another rework of the core combinePixels function (moving the switch on the number of valid inputs inside the iteration). Much happier with the outlier rejection now.

File:

• branches/pap/psModules/src/imcombine/pmStack.h (modified) (1 diff)

branches/pap/psModules/src/imcombine/pmStack.h

@@ -45 +45 @@
                     psArray *input,     ///< Input array of pmStackData
                     psImageMaskType maskVal, ///< Mask value of bad pixels
+                    psImageMaskType suspect, ///< Mask value of suspect pixels
                     psImageMaskType bad,     ///< Mask value to give rejected pixels
                     int kernelSize,     ///< Half-size of the convolution kernel
-                    int numIter,        ///< Number of iterations
+                    float iter,         ///< Number of iterations per input
                     float rej,          ///< Rejection limit (standard deviations)
                     float sys,          ///< Relative systematic error