adaptive -- Adaptive filter for an image.
adaptive input output
COPYRIGHT (c) 1990 European Southern Observatory IDENT filter.adap.hlq AUTHOR G.M. Richter KEYWORDS MIDAS, Smoothing, gradient-filter, Laplace-filter PURPOSE On-line help file for the command: FILTER/ADAPTIV VERSION 1.0 10-APR-1991 : Creation, Richter
This task evalutes---using an H-transform---the local signal-to-noise ratio as a function of decreasing resolution: mean gradients and curvatures over different scale lengths (obtained from the H-coefficients of different orders) are compared to the corresponding expectation values of the noise. The order for which this signal-to-noise ratio exceeds a given parameter threshold indicates the local resolution scale length of the signal (i.e., the point becomes significant at this order), and this determines the size of the impulse response of the filter at this point.
When the task is completed, it displays information about the standard deviation and expectation values of the gradients and the Laplace terms at every order involved, and the number of pixels that became significant on every order by the gradient and by the Laplace term, respectively. The rest of the pixels are set to the maximum size response.
- input = "" [file name]
- Name of input image.
- output = "" [file name]
- Name of output image.
- (mask = "") [file name]
- Name of optional mask. If mask = "", no mask is used. If a mask is used, it must be the same size as the input image. A given pixel is included when estimating noise statistics if its corresponding mask value is zero, or if no mask is used. Any non-zero mask value means the image pixel will not be used for noise statistics.
- (type = "smooth") [string, allowed values: smooth | gradient | laplace]
- The filter type may be smoothing, gradient filter, or Laplace filter.
- (shape = "pyramid") [string, allowed values: box | pyramid]
- The impulse response shape may be either box or pyramid. Pyramid is recommended.
- (size = 31) [integer, min=1, max=INDEF]
- Maximum size of the impulse response filter. In regions of high resolution the actual size is smaller. Possible sizes are: for box: 3,5,9,17,33,65,129 pixels; for pyramid: 3,5,7,11,15,23,31,47,63,95,127.
- (threshold = 3.) [real]
- Threshold for significance (see description above).
- (noise = "additive") [string, allowed values: additive | poisson]
- Either an additive or Poisson noise model may be assumed.
1. The image "frame" is adaptively smoothed with a maximum filter size of 31x31 pixels. The noise statistics are estimated from the whole image rather than from a mask (signal is included so care must be used when the image has a strong signal!) The output image is "result".
pl> adaptive frame result "" s p 31 3 a