Auxilliary Tool: A class for efficiently applying multiple box convolutions to the same image.
More...
#include <OpenCV2BoxFilter.h>
Public Member Functions | |
| OpenCV2BoxFilter (int maxPaddingRequired, int filteType=CV_32F) | |
| Constructor. | |
| OpenCV2BoxFilter () | |
| Constructor. | |
| OpenCV2BoxFilter (const OpenCV2BoxFilter &rhs) | |
| Copy Constructor. | |
| OpenCV2BoxFilter & | operator= (const OpenCV2BoxFilter &rhs) |
| Assignment Operator. | |
| ~OpenCV2BoxFilter () | |
| Default Destructor. | |
| void | setNewImage (const cv::Mat &imageToFilter) |
| Change the image that will be filtered. The source image is assumed to be single channel image, of a type supported by cv::integral. | |
| void | setBoxFilter (cv::Mat &destinationImage, cv::Rect boxPosition=cv::Rect(0, 0, 1, 1), double scaleResultFactor=1.0) |
| Compute a box filter and over-write the result to an image. The result of filtering is the sum of the pixel values in the specified rectangular region. It will be a 64-bit single channel floating point image. | |
| void | accumulateBoxFilter (cv::Mat &destinationImage, cv::Rect boxPosition=cv::Rect(0, 0, 1, 1), double scaleResultFactor=1.0) |
| Compute a box filter and add the result to an image. The result of filtering is the average of the pixel values in the specified rectangular region. This is useful for designing filters composed of multiple boxes. | |
Auxilliary Tool: A class for efficiently applying multiple box convolutions to the same image.
By computing the integral image once, an image can be filtered multiple times with boxes of different sizes. Requires OpenCV libraries.
| OpenCV2BoxFilter::OpenCV2BoxFilter | ( | int | maxPaddingRequired, |
| int | filteType = CV_32F |
||
| ) |
Constructor.
NOTE: The source image is assumed to be single channel image, of a type supported by cv::Integral. The destination image will be a 64-bit floating point image.
| maxPaddingRequired | Extra memory required for performing convolutions. In practice, this must be the distance from the central pixel of the most extreme point of any box we expect to use, i.e. at least the maximum intended value of: (1) |boxPosition.x|, (2) |boxPosition.x+boxPosition.width-1|, (3) |boxPosition.y|, (4) |boxPosition.y+boxPosition.height-1|. Erring on the side of too much padding has no adverse consequences other than increased memory usage. |
| filterType | OpenCV data type for output of filtering. Can be CV_32S, CV_32F, CV_64F. |
| OpenCV2BoxFilter::OpenCV2BoxFilter | ( | ) |
Constructor.
NOTE: The source image is assumed to be single channel image, of a type supported by cv::Integral. The destination image will be a 64-bit floating point image.
NOTE: call setMaxPaddingRequired later.
| OpenCV2BoxFilter::~OpenCV2BoxFilter | ( | ) |
Default Destructor.
Deallocates all memory associated with the box filter.
| void OpenCV2BoxFilter::accumulateBoxFilter | ( | cv::Mat & | destinationImage, |
| cv::Rect | boxPosition = cv::Rect(0, 0, 1, 1), |
||
| double | scaleResultFactor = 1.0 |
||
| ) |
Compute a box filter and add the result to an image. The result of filtering is the average of the pixel values in the specified rectangular region. This is useful for designing filters composed of multiple boxes.
| destinationImage | Image target for the results of the filtering. After invoking this method, the contents of this variable will be overwritten with the results of the operation. |
| boxPosition | Location of the box (relative to the central pixel) to average pixel values over. For example, cvRect(0,0,1,1) has zero-offset from the current pixel and a width/height of one pixel. Filtering the image in this way will return the original image. To filter with a width-5 averaging filter with no offset, use cvRect(-2,-2,5,5). |
| scaleResultFactor | Scale the result by some factor. For example, if the scale is boxPosition.width*boxPosition.height, the sum of the pixels in the box is returned instead of the average. |
| void OpenCV2BoxFilter::setBoxFilter | ( | cv::Mat & | destinationImage, |
| cv::Rect | boxPosition = cv::Rect(0, 0, 1, 1), |
||
| double | scaleResultFactor = 1.0 |
||
| ) |
Compute a box filter and over-write the result to an image. The result of filtering is the sum of the pixel values in the specified rectangular region. It will be a 64-bit single channel floating point image.
| destinationImage | Image target for the results of the filtering. After invoking this method, the contents of this variable will be overwritten with the results of the operation. |
| boxPosition | Location of the box (relative to the central pixel) to average pixel values over. For example, cvRect(0,0,1,1) has zero-offset from the current pixel and a width/height of one pixel. Filtering the image in this way will return the original image. To filter with a width-5 averaging filter with no offset, use cvRect(-2,-2,5,5). |
| scaleResultFactor | Scale the result by some factor. For example, if the scale is 1/boxPosition.width*boxPosition.height, the average of the pixels in the box is returned instead of the sum. |
| void OpenCV2BoxFilter::setNewImage | ( | const cv::Mat & | imageToFilter ) |
Change the image that will be filtered. The source image is assumed to be single channel image, of a type supported by cv::integral.
This method computes a new integral image that will be used repeatedly for future filtering, until this method is invoked again with a new image.
| imageToFilter | A new OpenCV image to filter repeatedly. |
1.7.2