Standard Noise and Texture Filters.

• thres - threshold TOS at making all pixels less than IP(thres) zero, operates only on real component of complex images.
• lsf - linear sequential (exponential) smoothing filter of TOS with decay parameter and up/down left right handedness given by IP(sigma,LR,UD). Useful for fast large scale spatial smoothing. Two applications give a reasonable approximation to a Gaussian (ie: Central Limit Theorem).
• med - median filter of TOS for 3x3 pixel patch. An edge preserving noise filter which is particularly effective at removing salt and pepper (drop-out) noise. Noise is reduced by a (typical) factor of two and image quantization is unaffected.
• rank - rank order filtering of TOS with pixel region IP(range) and expected grey level noise level of IP(const). Useful for allowing parametric statistical measures to be applied to non-parametric image data. The resulting image has approximately bi-nomial noise characteristics which can be transformed to a domain of random uniform errors via the asin mapping.
• gauss - gaussian filter of TOS image with standard deviation and convolution area of IP( sigma, range). An optimally compact (spatial/frequency) noise filter. Generally used for band-pass filtering and to reduce noise (particularly the effects of data quantization) before derivative estimation. Errors are generally reduced by an (approximate) factor of two.
• tsmooth - local smoothed average of TOS image along direction of minimum gradient. An edge preserving noise filter which reduces image noise by a factor of (approximately) $\sqrt 2$ (see Tina memo 2003-007).