Box Image Filter

Overview

Box Image Filter is a low-pass filter that smooths the image by making each output pixel the average of the surrounding ones, removing details, noise and and edges from images.

Input	Window size	Output
	5x5

Implementation

Box image filter is implemented as a convolution operation on the input image using the following kernel:

\[ box_{m,n} = \frac{1}{mn} \begin{bmatrix} 1 & 1 & \dots & 1 \\ 1 & 1 & \dots & 1 \\ \vdots & \vdots & \ddots & \vdots \\ 1 & 1 & \dots & 1 \end{bmatrix}_{m \times n} \]

Usage

1. Initialization phase
   1. Include the header that defines the box filter function.

      #include <vpi/algo/BoxImageFilter.h>

   2. Define the stream on which the algorithm will be executed, the input and output images.

          VPIStream stream = /*...*/;
          VPIImage input   = /*...*/;

   3. Create the output image.

          uint32_t w, h;
          vpiImageGetSize(input, &w, &h);
       
          VPIImageType type;
          vpiImageGetType(input, &type);
       
          VPIImage output;
          vpiImageCreate(w, h, type, 0, &output);

2. Processing phase
   1. Submit the algorithm to the stream, input, output images, window size and boundary condition.

          vpiSubmitBoxImageFilter(stream, input, output, 5, 5, VPI_BOUNDARY_COND_ZERO);

   2. Optionally, wait until the processing is done.

          vpiStreamSync(stream);

Limitations and Constraints

Constraints for specific backends supersede the ones specified for all backends.

All Backends

• Kernel size must be at least 1x1 and at most 11x11.
• Accepted boundary conditions:
  • VPI_BOUNDARY_COND_CLAMP
  • VPI_BOUNDARY_COND_ZERO
• Input and output images must have same image type and size.
• Accepted input image type:
  • VPI_IMAGE_TYPE_U8
  • VPI_IMAGE_TYPE_S8
  • VPI_IMAGE_TYPE_U16
  • VPI_IMAGE_TYPE_S16

PVA

• Minimum kernel size is 2x2.
• Accepted boundary conditions:
  • VPI_BOUNDARY_COND_ZERO
  • VPI_BOUNDARY_COND_CLAMP only if kernel size is 3x3.

Performance

For further information on how performance was benchmarked, see Performance Measurement.

Jetson AGX Xavier

size	type	kernel	boundary	CPU	CUDA	PVA
1920x1080	u8	3x3	zero	0.4 ms	0.0647 ms	0.9614 ms
1920x1080	u8	3x3	clamp	0.675 ms	0.0637 ms	1.0782 ms
1920x1080	u8	5x5	zero	0.43 ms	0.0690 ms	1.2316 ms
1920x1080	u8	7x7	zero	0.98 ms	0.0881 ms	1.7586 ms
1920x1080	u8	11x11	zero	1.30 ms	0.0994 ms	3.2608 ms
1920x1080	u16	3x3	zero	0.839 ms	0.1075 ms	1.0610 ms
1920x1080	u16	3x3	clamp	0.815 ms	0.1071 ms	1.0676 ms
1920x1080	u16	5x5	zero	1.3 ms	0.1153 ms	1.5123 ms
1920x1080	u16	7x7	zero	1.08 ms	0.1353 ms	2.3511 ms
1920x1080	u16	11x11	zero	1.284 ms	0.1582 ms	4.6477 ms

Jetson TX2

size	type	kernel	boundary	CPU	CUDA	PVA
1920x1080	u8	3x3	zero	1.5 ms	0.255 ms	n/a
1920x1080	u8	3x3	clamp	1.41 ms	0.255 ms	n/a
1920x1080	u8	5x5	zero	1.72 ms	0.288 ms	n/a
1920x1080	u8	7x7	zero	2.29 ms	0.395 ms	n/a
1920x1080	u8	11x11	zero	2.99 ms	0.472 ms	n/a
1920x1080	u16	3x3	zero	1.86 ms	0.380 ms	n/a
1920x1080	u16	3x3	clamp	1.84 ms	0.377 ms	n/a
1920x1080	u16	5x5	zero	2.2 ms	0.423 ms	n/a
1920x1080	u16	7x7	zero	2.55 ms	0.578 ms	n/a
1920x1080	u16	11x11	zero	3.37 ms	0.677 ms	n/a

Jetson Nano

size	type	kernel	boundary	CPU	CUDA	PVA
1920x1080	u8	3x3	zero	3.00 ms	0.6692 ms	n/a
1920x1080	u8	3x3	clamp	2.995 ms	0.6591 ms	n/a
1920x1080	u8	5x5	zero	3.792 ms	0.7444 ms	n/a
1920x1080	u8	7x7	zero	4.76 ms	1.025 ms	n/a
1920x1080	u8	11x11	zero	6.77 ms	1.237 ms	n/a
1920x1080	u16	3x3	zero	3.550 ms	0.964 ms	n/a
1920x1080	u16	3x3	clamp	3.56 ms	0.969 ms	n/a
1920x1080	u16	5x5	zero	4.21 ms	1.018 ms	n/a
1920x1080	u16	7x7	zero	5.305 ms	1.386 ms	n/a
1920x1080	u16	11x11	zero	7.44 ms	1.657 ms	n/a