Low Pass Filter

Functions
I32	gauss (image *src, image *dst, F32 sigma)
	Gauss Filter (Recursive Version). More...
I32	gauss_hor (image *src, image *dst, F32 sigma)
	Horizontal Gauss Filter (Recursive Version). More...
I32	gauss_ver (image *src, image *dst, F32 sigma)
	Vertical Gauss Filter (Recursive Version). More...
I32	gauss_fir (image *src, image *dst, F32 sigma)
	Non-Recursive Gauss Filter (FIR). More...
I32	boxavg (image *src, image *dst, I32 kx, I32 ky)
	Averaging Over a Rectangle. More...
void	ii_boxavg (image *integral, image *dst, I32 kx, I32 ky)
	Integral Image Based Averaging Over a Rectangle. More...
I32	avg (image *a, image *b, I32 kx, I32 ky, void(*func)(), I32 v)
	Moving Average or Unsharp Masking of an Image Variable. More...
I32	avg2 (image *a, image *b, I32 kx, I32 ky, void(*func)(), I32 v)
	Moving Average or Unsharp Masking of an Image Variable. More...
I32	isef (image *src, image *dst, F32 b)
	Infinite Symmetric Exponential Filter (Recursive Version). More...
I32	isef_hor (image *src, image *dst, F32 b)
	Horizontal Infinite Symmetric Exponential Filter (Recursive Version). More...
I32	isef_ver (image *src, image *dst, F32 b)
	Vertical Infinite Symmetric Exponential Filter (Recursive Version). More...
#define	avgm(a, b, kx, ky)   avg(a, b, kx, ky, (void (*)())0, 0)
	Moving Average. More...
#define	maskx(a, b, kx, ky, v)   avg(a, b, kx, ky, (void (*)())FL_sub2x, v)
	Unsharp Masking of the Image. More...
#define	masky(a, b, kx, ky)   avg(a, b, kx, ky, (void (*)())FL_sub2y, 0)
	Unsharp Masking of the Image. More...
#define	avgm2(a, b, kx, ky)   avg2(a, b, kx, ky, (void (*)())0, 0)
	Moving Average. More...
#define	maskx2(a, b, kx, ky, v)   avg2(a, b, kx, ky, (void (*)())FL_sub2x, v)
	Unsharp Masking of the Image. More...
#define	masky2(a, b, kx, ky)   avg2(a, b, kx, ky, (void (*)())FL_sub2y, 0)
	Unsharp Masking of the Image. More...

Detailed Description

Macro Definition Documentation

avgm

#define avgm	(		a,
			b,
			kx,
			ky
	)		avg(a, b, kx, ky, (void (*)())0, 0)

The macro calculates the moving average filter of image variable a and stores the result in image variable b. kx and ky are the horizontal and vertical kernel size. More information at avg().

The result image will be centered according to the kernel size (kx, ky), i.e. the (smaller) result image will start at location b->st + kx/2 + (ky/2) * b->pitch.

avgm() is a macro which calls avg() with basic function void(*)() 0 as an argument.

maskx

#define maskx	(		a,
			b,
			kx,
			ky,
			v
	)		avg(a, b, kx, ky, (void (*)())FL_sub2x, v)

The macro calculates a so-called unsharp masking of image a to image b. kx and ky are the horizontal and vertical kernel size used to blur the image (moving average). The offs raises the resulting grey value given by that formula: b = clip(0,255, [a - avgm(a) + offs]). More information at avg().

The result image will be centered according to the kernel size (kx, ky), i.e. the (smaller) result image will start at location b->st + kx/2 + (ky/2) * b->pitch.

maskx() is a macro which calls avg() with basic function FL_sub2x() as an argument.

masky

#define masky	(		a,
			b,
			kx,
			ky
	)		avg(a, b, kx, ky, (void (*)())FL_sub2y, 0)

The macro calculates a so-called unsharp masking of image a and binarizes it to the image b. kx and ky are the horizontal and vertical kernel size used to blur the image (moving average). The output will be given according to that formula: b = ([a - avgm(a)]>0)?(255):(0). More information at avg().

The result image will be centered according to the kernel size (kx, ky), i.e. the (smaller) result image will start at location b->st + kx/2 + (ky/2) * b->pitch.

masky() is a macro which calls avg() with basic function FL_sub2y() as an argument.

avgm2

#define avgm2	(		a,
			b,
			kx,
			ky
	)		avg2(a, b, kx, ky, (void (*)())0, 0)

The macro calculates the moving average filter of image variable a and stores the result in image variable b. kx and ky are the horizontal and vertical kernel size. More information at avg2().

The result will be placed in the left upper corner of b.

avgm2() is a macro which calls avg2() with basic function void(*)() 0 as an argument.

maskx2

#define maskx2	(		a,
			b,
			kx,
			ky,
			v
	)		avg2(a, b, kx, ky, (void (*)())FL_sub2x, v)

The macro calculates a so-called unsharp masking of image a to image b. kx and ky are the horizontal and vertical kernel size used to blur the image (moving average). The offs raises the resulting grey value given by that formula: b = clip(0,255, [a - avgm2(a) + offs]). More information at avg2().

The result will be placed in the left upper corner of b.

maskx2() is a macro which calls avg2() with basic function FL_sub2x() as an argument.

masky2

#define masky2	(		a,
			b,
			kx,
			ky
	)		avg2(a, b, kx, ky, (void (*)())FL_sub2y, 0)

The macro calculates a so-called unsharp masking of image a and binarizes it to the image b. kx and ky are the horizontal and vertical kernel size used to blur the image (moving average). The output will be given according to that formula: b = ([a - avgm2(a)]>0)?(255):(0). More information at avg2().

The result will be placed in the left upper corner of b.

masky2() is a macro which calls avg2() with basic function FL_sub2y() as an argument.

Function Documentation

gauss()

I32 gauss	(	image *	src,
		image *	dst,
		float	sigma
	)

The function calculates the recursive gauss filter with filter parameter sigma with 0.0 ≤ sigma ≤ 5.0. sigma defines the equivalent of the filter kernel size (standard deviation) for this recursive filter: The larger the value of sigma, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of sigma.

Remarks

The function calls gauss_hor() and gauss_ver() in succession.

Memory Consumption

2*(dx*(dy+6)+1) Bytes of Heap Memory.

See also

gauss_fir(), gauss_hor(), gauss_ver().

Parameters

src	Source Image.
dst	Destination Image.
sigma	Filter Parameter (0.0 ≤ sigma ≤ 5.0).

Point spread funtions for different values of sigma (values in Hex):

• sigma = 0.625:

  

  .
• sigma = 1.0:

  

  .
• sigma = 1.5:

  

  .

Return values

ERR_MEMORY	if Memory Allocation Fails.
ERR_PARAM	if sigma is not in Range [0.0,5.0].
ERR_NONE	on Success.

gauss_hor()

I32 gauss_hor	(	image *	src,
		image *	dst,
		float	sigma
	)

The function calculates the horizontal recursive gauss filter with filter parameter sigma with 0.0 ≤ sigma ≤ 5.0. sigma defines the equivalent of the filter kernel size (standard deviation) for this recursive filter: The larger the value of sigma, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of sigma.

Parameters

src	Source image.
dst	Destination image.
sigma	Filter Parameter (0.0 ≤ sigma ≤ 5.0).

Memory Consumption

2*(dx+1) Bytes of Heap Memory.

See also

gauss(), gauss_ver().

Return values

ERR_PARAM	if sigma is not in Range [0.0,5.0].
ERR_MEMORY	if Memory Allocation Fails.
ERR_NONE	on Success.

gauss_ver()

I32 gauss_ver	(	image *	src,
		image *	dst,
		float	sigma
	)

The function calculates the vertical recursive gauss filter with filter parameter sigma with 0.0 ≤ sigma ≤ 5.0. sigma defines the equivalent of the filter kernel size (standard deviation) for this recursive filter: the larger the value of sigma, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of sigma.

Parameters

src	Source Image.
dst	Destination Image.
sigma	Filter Parameter.

Memory Consumption

2*(dx*(dy+6)+1) Bytes of Heap Memory.

See also

gauss(), gauss_hor().

Return values

ERR_PARAM	if sigma is not in Range [0.0,5.0].
ERR_MEMORY	if Memory Allocation Fails.
ERR_NONE	on Success.

gauss_fir()

I32 gauss_fir	(	image *	src,
		image *	dst,
		float	sigma
	)

This is the non-recursive version of the gauss low-pass filter (FIR). sigma is the standard deviation of the filter.

sigma	Filter Size
0.391	3x3
0.625	5x5
0.812	7x7

Values for sigma between the values in the table are allowed. The function switches to whatever filter size comes closer to the value of sigma. The function returns the standard error code.

Parameters

src	Source Image.
dst	Destination.
sigma	Standard Deviation for Gauss Filter.

See also

gauss().

Return values

ERR_TYPE	if src or dst Image is no Grey Image.
ERR_NONE	on Success.

boxavg()

I32 boxavg	(	image *	src,
		image *	dst,
		I32	kx,
		I32	ky
	)

This function outputs the arithmetic average of values of pixels lying in a box of dimension (kx,ky) with its output pixel at the center if kx and ky are odd. Data near the image border is extrapolated by repeating the border pixel values perpendicular to the orientation of the border.

Parameters

src	Source Image of Type IMAGE_GREY.
dst	Destination Image of Type IMAGE_GREY.
kx	Horizontal Averaging Pixel Column Count.
ky	Horizontal Averaging Pixel Row Count.

See also

avgm().

Return values

ERR_MEMORY	if Memory Allocation Fails.
integral_image()	if not ERR_NONE.
ERR_NONE	on Success.

ii_boxavg()

void ii_boxavg	(	image *	integral,
		image *	dst,
		I32	kx,
		I32	ky
	)

The arithmetic mean will be calculated using Integral Image Data. The integral image must have the dimensions ( dst->dx+kx, dst->dy+ky) for kx and ky odd. Data is aligned the following way: The pixel at dst(0,0) will contain the mean value over the (1,1)-(kx,ky) square of pixels in the image which was integrated (Average at Integral Image is (0,0)+(kx,ky)-(kx,0)-(0,ky), See Integral Image Theory). You do want to use this function with ImageAllocateEdge().

No validity checks are done!

Parameters

integral	Source Integral Image of Type IMAGE_GREY32.
dst	Destination Image of Type IMAGE_GREY.
kx	Horizontal Averaging Pixel Column Count.
ky	Horizontal Averaging Pixel Row Count.

Memory Consumption

None.

avg()

I32 avg	(	image *	a,
		image *	b,
		I32	kx,
		I32	ky,
		void(*)()	func,
		I32	v
	)

The function calculates the moving average filter of image variable a and stores the result in image variable b. The size of the moving average is specified with the values kx (horizontal kernel size) and ky (vertical kernel size). Images specified by image variables a and b must be different. If func is NULL, the function will calculate the moving average. If a function address is given, the original image will be subtracted from the moving average, performing an 'unsharp masking' operation.

There is a partner function avg2() which performs the same task using a different output alignment.

For avg() the result image will be centered according to the kernel size (kx, ky), i.e. the (smaller) result image will start at location b->st + kx/2 + (ky/2) * b->pitch.

The function pointer passed specifies the type of subtraction being performed.

The return value is negative, if an error is encountered. The following macros are available:

Macro Call	Result	func
avgm()	Moving Average	void (*)()0
maskx()	b = (a - avg + offs); clipping if c>255 or c<0	FL_sub2x()
masky()	b = (a - avg)>0 ? 255 : 0	FL_sub2y()

Of course, you can write your own subtract functions. Pass their address (function pointer) to avg().

Memory Consumption

8 * (dx/2 + 1) Bytes of Heap Memory.

See also

avg2().

ff3(), ff5().

Return values

ERR_FORMAT	if b->dx/y < a->dx/y - kx/y + 1.
ERR_MEMORY	if Memory Allocation failed.
ERR_NONE	on Success.

avg2()

I32 avg2	(	image *	a,
		image *	b,
		I32	kx,
		I32	ky,
		void(*)()	func,
		I32	v
	)

The function avg() calculates the moving average filter of image variable a and stores the result in image variable b. The size of the moving average is specified with the values kx (horizontal kernel size) and ky (vertical kernel size). Images specified by image variables a and b must be different. If func is NULL, the function will calculate the moving average. If a function address is given, the original image will be subtracted from the moving average, performing an 'unsharp masking' operation.

There is a partner function avg() which performs the same task using a different output alignment.

For avg2() the result will be placed in the left upper corner of b.

The function pointer passed specifies the type of subtraction being performed.

The return value is negative, if an error is encountered. The following macros are available:

Macro Call	Result	func
avgm2()	Moving Average	void (*)()0
maskx2()	b = (a - avg + offs); clipping if c>255 or c<0	FL_sub2x()
masky2()	b = (a - avg)>0 ? 255 : 0	FL_sub2y()

Of course, you can write your own subtract functions. Pass their address (function pointer) to avg2().

Memory Consumption

8 * (dx/2 + 1) Bytes of Heap Memory.

See also

avg().

ff3(), ff5().

Return values

ERR_FORMAT	if b->dx/y < a->dx/y - kx/y + 1.
ERR_MEMORY	if Memory Allocation failed.
ERR_NONE	on Success.

isef()

I32 isef	(	image *	src,
		image *	dst,
		F32	b
	)

The function calculates the infinite symmetric exponential filter with filter parameter b with 0.0 < b < 1.0. b defines the equivalent of the filter kernel size for this recursive filter: the larger the value of b, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of b.

Parameters

src	Source Image.
dst	Destination Image.
b	Filter Parameter.

Remarks

The function calls isef_hor() and isef_ver() in succession.

Memory Consumption

(dx*(dy+2)+2)/2 Bytes of Heap Memory.

See also

isef_hor(), isef_ver().

Return values

ERR_MEMORY	if Memory Allocation Fails.
ERR_TYPE	if src or dst Image is no Grey Image.
ERR_PARAM	if b is not in Range [0.0,1.0].
ERR_NONE	on Success.

isef_hor()

I32 isef_hor	(	image *	src,
		image *	dst,
		F32	b
	)

The function calculates the horizontal infinite symmetric exponential filter with filter parameter b with 0.0 < b < 1.0. b defines the equivalent of the filter kernel size for this recursive filter: the larger the value of b, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of b.

Parameters

src	Source Image.
dst	Destination Image.
b	Filter Parameter.

Memory Consumption

2*(dx+1) Bytes of Heap Memory.

See also

isef(), isef_ver().

Return values

ERR_TYPE	if src or dst Image is no Grey Image.
ERR_PARAM	if b is not in Range [0.0,1.0].
ERR_MEMORY	if Memory Allocation Fails.
ERR_NONE	on Success.

isef_ver()

I32 isef_ver	(	image *	src,
		image *	dst,
		F32	b
	)

The function calculates the vertical infinite symmetric exponential filter with filter parameter b with 0.0 < b < 1.0. b defines the equivalent of the filter kernel size for this recursive filter: the larger the value of b, the larger the kernel size. Since this function is designed as a recursive filter, the execution speed does not depend on the size of b.

Parameters

src	Source Image.
dst	Destination Image.
b	Filter Parameter.

Memory Consumption

2*(dx*(dy+2)+2) Bytes of Heap Memory.

See also

isef(), isef_hor().

Return values

ERR_TYPE	if src or dst Image is no Grey Image.
ERR_PARAM	if b is not in Range [0.0,1.0].
ERR_MEMORY	if Memory Allocation Fails.
ERR_NONE	on Success.