HOME GUIDE OPERATIONS DOCS ERRORS FORMATS INSTALL NEW TIPS WEB SITES

BC - Box Convolution

(9/7/94)

PURPOSE

Performs various contrast enhancement operations in 2D and 3D using local averaging.

USAGE

.OPERATION: BC

.INPUT FILE: H_PIC001
[Enter name of image file to be enhanced.]

.OUTPUT FILE: OUT001
[Enter the name of the file where enhanced image is to be put. Sample Outputs : 1:(H)IGH PASS- H_IMG010, 2:(L)OW PASS - H_IMG011, 3:(T)HRES - H_IMG012, 4: (D)IVIS - H_IMG013]

(H)IGH PASS / (L)OW PASS / (T)HRES / (D)IVIS: H/L/T/D [Enter enhancement option:

Option "H" - High-Pass Filtering. For each point (IN) of the input image, the average (AV) of the surrounding area within a box (see below) is computed. The output image (OUT) is computed according to the following formula:

OUT = (IN-AV)*F+IN*(1-F)

where F is a weighting factor (see below). F=1.0 results in a pure high-pass filtered image. Values of F between 1.0 and 0.0 result in images with various degrees of high resolution enhancement.

Option "L" - Low-Pass Filtering. For each point of the input image, the average (AV) of a surrounding area within a box (see below) is computed. The output image is computed according to the following formula:

OUT = AV*F+IN*(1-F)

where F is a weighting factor (see below). F=1.0 results in a pure low-pass filtered image. Values of F between 1.0 and 0.0 result in images with various degrees of low spatial frequency enhancement.

Option "T" - Local Threshold Filtering (restricted to images that have been previously stretched or histogram equalized). For each point of the input image, the average (AV) of a surrounding area is computed. The output image is computed according to the following formula:

OUT = (1.-F)*IN FOR IN < AV OUT = (1.-F)*IN+F*2. FOR IN >= AV

where F is a weighting factor (see below). F=1.0 results in a purely threshold filtered image. Values of F between 1.0 and 0.0 result in images with various proportions of the input image added on.

Option "D" - Divisive Local Contrast Correction. For each point of the input image, the average (AV) of a surrounding area within a box (see below) is computed. Each point is divided by its local average. The formula for this procedure is:

OUT = F*IN/(AV+EPS)+FC*IN

where F is a weighting factor (see below) and EPS is an offset constant for avoiding zero division. (EPS is set to .1 times the global input average). F=1.0 results in a pure divisive-corrected image. Values of F between 1.0 and 0.0 result in images with various degrees of divisive correction.]

.LOCAL AREA SIZE COL,ROWS: 5,6
[Enter dimensions of averaging box. Note that the dimensions may be different. If only NSAM is entered, NROW=NSAM will be assumed.]

If the input is a volume the following question appears:

.LOCAL AREA SIZE SLICES: 5,6
[Enter dimensions of averaging box. Note that the dimensions may be different. If only NSAM is entered, NROW=NSAM will be assumed.]

FILTER WEIGHT (0.0->1.0): 0.8 [Enter weight for mixing]

NOTES

  1. The margins are replaced by the nearest value possible to calculate. This can result in errors for structures filling the image close to boundaries and for large local area (box) dimensions.

SUBROUTINES

BOXX, AVERG, AVERG3

CALLER

UTIL3

© Copyright Notice /       Enquiries: spider@wadsworth.org